diff --git a/.gitattributes b/.gitattributes
index bc3796a4bc551e4e027dcfb0bc32b0884b04723a..77a3fdd4861d196cfde7e7973cf627c1fdde206a 100644
--- a/.gitattributes
+++ b/.gitattributes
@@ -60,3 +60,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
 modelblocks_kenlm_models/openwebtext.1.fw.kenlm filter=lfs diff=lfs merge=lfs -text
 owt/unigram_counts.tsv filter=lfs diff=lfs merge=lfs -text
+verification/wave1-tokens.json filter=lfs diff=lfs merge=lfs -text
+verification/wave2.json filter=lfs diff=lfs merge=lfs -text
diff --git a/README.md b/README.md
index a239776c977fe209ba9d827fa0a28eb11c3e2830..2eb8f00aee5fe94c04ea78ad8577969277ebad6d 100644
--- a/README.md
+++ b/README.md
@@ -1,5 +1,5 @@
 ---
-pretty_name: Per-word reading predictors
+pretty_name: REPLICA predictor store
 tags:
   - psycholinguistics
   - reading-times
@@ -8,6 +8,8 @@ tags:
 configs:
   - config_name: tags
     data_files: tags.parquet
+  - config_name: coverage
+    data_files: coverage.parquet
   - config_name: brothers_kuperberg
     data_files: brothers_kuperberg.parquet
   - config_name: brown_spr
@@ -16,6 +18,20 @@ configs:
     data_files: bsc.parquet
   - config_name: celer
     data_files: celer.parquet
+  - config_name: characters_celer
+    data_files: characters/celer.parquet
+  - config_name: characters_geco
+    data_files: characters/geco.parquet
+  - config_name: characters_meco_l1_w2_ch_s
+    data_files: characters/meco_l1_w2_ch_s.parquet
+  - config_name: characters_meco_l1_w2_ch_t
+    data_files: characters/meco_l1_w2_ch_t.parquet
+  - config_name: characters_provo
+    data_files: characters/provo.parquet
+  - config_name: characters_sbsat
+    data_files: characters/sbsat.parquet
+  - config_name: characters_zuco1
+    data_files: characters/zuco1.parquet
   - config_name: copco
     data_files: copco.parquet
   - config_name: devarda2023
@@ -30,30 +46,6 @@ configs:
     data_files: geco.parquet
   - config_name: hubbard2019_n400
     data_files: hubbard2019_n400.parquet
-  - config_name: meco_char_de
-    data_files: meco_char_de.parquet
-  - config_name: meco_char_du
-    data_files: meco_char_du.parquet
-  - config_name: meco_char_ee
-    data_files: meco_char_ee.parquet
-  - config_name: meco_char_en
-    data_files: meco_char_en.parquet
-  - config_name: meco_char_fi
-    data_files: meco_char_fi.parquet
-  - config_name: meco_char_he
-    data_files: meco_char_he.parquet
-  - config_name: meco_char_it
-    data_files: meco_char_it.parquet
-  - config_name: meco_char_ko
-    data_files: meco_char_ko.parquet
-  - config_name: meco_char_no
-    data_files: meco_char_no.parquet
-  - config_name: meco_char_ru
-    data_files: meco_char_ru.parquet
-  - config_name: meco_char_sp
-    data_files: meco_char_sp.parquet
-  - config_name: meco_char_tr
-    data_files: meco_char_tr.parquet
   - config_name: meco_de
     data_files: meco_de.parquet
   - config_name: meco_du
@@ -136,6 +128,76 @@ configs:
     data_files: szewczyk_federmeier2022_n400.parquet
   - config_name: szewczyk_n400
     data_files: szewczyk_n400.parquet
+  - config_name: tokens_celer
+    data_files: tokens/celer.parquet
+  - config_name: tokens_copco
+    data_files: tokens/copco.parquet
+  - config_name: tokens_dundee
+    data_files: tokens/dundee.parquet
+  - config_name: tokens_federmeier2007_n400
+    data_files: tokens/federmeier2007_n400.parquet
+  - config_name: tokens_geco
+    data_files: tokens/geco.parquet
+  - config_name: tokens_meco_de
+    data_files: tokens/meco_de.parquet
+  - config_name: tokens_meco_ee
+    data_files: tokens/meco_ee.parquet
+  - config_name: tokens_meco_fi
+    data_files: tokens/meco_fi.parquet
+  - config_name: tokens_meco_gr
+    data_files: tokens/meco_gr.parquet
+  - config_name: tokens_meco_he
+    data_files: tokens/meco_he.parquet
+  - config_name: tokens_meco_it
+    data_files: tokens/meco_it.parquet
+  - config_name: tokens_meco_ko
+    data_files: tokens/meco_ko.parquet
+  - config_name: tokens_meco_l1_w2_ba
+    data_files: tokens/meco_l1_w2_ba.parquet
+  - config_name: tokens_meco_l1_w2_bp
+    data_files: tokens/meco_l1_w2_bp.parquet
+  - config_name: tokens_meco_l1_w2_ch_s
+    data_files: tokens/meco_l1_w2_ch_s.parquet
+  - config_name: tokens_meco_l1_w2_ch_t
+    data_files: tokens/meco_l1_w2_ch_t.parquet
+  - config_name: tokens_meco_l1_w2_ge_po
+    data_files: tokens/meco_l1_w2_ge_po.parquet
+  - config_name: tokens_meco_l1_w2_ge_zu
+    data_files: tokens/meco_l1_w2_ge_zu.parquet
+  - config_name: tokens_meco_l1_w2_hi_iiith
+    data_files: tokens/meco_l1_w2_hi_iiith.parquet
+  - config_name: tokens_meco_l1_w2_hi_iitk
+    data_files: tokens/meco_l1_w2_hi_iitk.parquet
+  - config_name: tokens_meco_l1_w2_ic
+    data_files: tokens/meco_l1_w2_ic.parquet
+  - config_name: tokens_meco_l1_w2_tr
+    data_files: tokens/meco_l1_w2_tr.parquet
+  - config_name: tokens_meco_l2_w2
+    data_files: tokens/meco_l2_w2.parquet
+  - config_name: tokens_meco_ru
+    data_files: tokens/meco_ru.parquet
+  - config_name: tokens_meco_tr
+    data_files: tokens/meco_tr.parquet
+  - config_name: tokens_natural_stories
+    data_files: tokens/natural_stories.parquet
+  - config_name: tokens_natural_stories_maze
+    data_files: tokens/natural_stories_maze.parquet
+  - config_name: tokens_onestop
+    data_files: tokens/onestop.parquet
+  - config_name: tokens_potec
+    data_files: tokens/potec.parquet
+  - config_name: tokens_provo
+    data_files: tokens/provo.parquet
+  - config_name: tokens_sbsat
+    data_files: tokens/sbsat.parquet
+  - config_name: tokens_szewczyk_federmeier2022_n400
+    data_files: tokens/szewczyk_federmeier2022_n400.parquet
+  - config_name: tokens_szewczyk_n400
+    data_files: tokens/szewczyk_n400.parquet
+  - config_name: tokens_wlotko_federmeier2012_n400
+    data_files: tokens/wlotko_federmeier2012_n400.parquet
+  - config_name: tokens_zuco1
+    data_files: tokens/zuco1.parquet
   - config_name: ucl_et
     data_files: ucl_et.parquet
   - config_name: ucl_spr
@@ -150,540 +212,250 @@ configs:
     data_files: zuco_n400.parquet
 ---
 
-# Per-word reading predictors
+# REPLICA predictor store
 
-Precomputed per-word predictors for computational models of human sentence
-processing: word surprisal, several kinds of entropy, and unigram surprisal,
-scored over reading corpora (eye tracking, self-paced reading, N400, maze).
+Per-unit predictors for computational models of human sentence processing:
+surprisal, several entropies, similarity-adjusted surprisal, information
+value and unigram surprisal, scored by language models over reading
+corpora (eye tracking, self-paced reading, maze, N400). Every row names the
+text unit it describes and the unit its number is in.
 
-Every row says what text unit it describes and what unit its number is in:
+| `dataset` | `tag` | `predictor` | `stimulus_id` | `unit_index` | `text_unit` | `value` | `value_unit` | `layer` | `storage_class` | `unit_rule` |
+|---|---|---|---|---|---|---|---|---|---|---|
+| provo | `gpt2_surprisal_provo` | `surprisal` | provo_1 | 4 | Apple | 8.4213 | bits | 1 | positional | |
 
-| `dataset` | `tag` | `predictor` | `stimulus_id` | `unit_index` | `text_unit` | `value` | `value_unit` |
-|---|---|---|---|---|---|---|---|
-| provo | `gpt2_ft_entropy_provo` | `token_entropy` | provo_1 | 4 | Apple | 8.4213 | bits |
+Start from `tags.parquet`: one row per (`tag`, `predictor`, file), with the
+model, method, configuration, storage class, layer and provenance of every
+cell decoded, so a cell can be found by model, method and configuration.
 
-Counts below describe cells indexed in `tags.parquet`. Cells are unique tags.
-Rows sum `n_rows` over (`tag`, `predictor`, file), so one word can contribute
-to several predictor rows.
+- 2,596 cells across 58 corpora, 280,030,555 scalar rows
+- 25 methods, 26 models and frequency tables
 
-<!-- TABLE:totals -->
-- **2,111 cells** across **58 corpora**, **142,164,743 scored rows**
-- 25 methods x 26 models x 158 predictor keys
-<!-- /TABLE:totals -->
+## Layout
 
-## Start here: `tags.parquet`
-
-One row per (`tag`, `predictor`, file), with the model, method and configuration
-decoded out of the tag, so you can find a cell by what it is instead of by
-guessing at its name.
+| path | contents |
+|---|---|
+| `tags.parquet` | the cell index: one row per (`tag`, `predictor`, file) |
+| `coverage.parquet` | one row per planned or produced table with its status |
+| `history.parquet` | first-publish and last-change commit of every file |
+| `<dataset>.parquet` | word-grain scalar predictors of one corpus |
+| `tokens/<dataset>.parquet` | token-grain scalar predictors of one corpus |
+| `characters/<dataset>.parquet` | character-grain scalar predictors of one corpus |
+| `responsivity/<tag>.parquet` | embedding-distance word predictors, one file per cell |
+| `lens_layers/<tag>.parquet` | per-layer log-probability vectors of the lens cells, one file per cell |
+| `mcword_samples/`, `cont_entropy_samples/` | per-unit Monte Carlo sample pools behind the entropy cells |
+| `next_log_probs/` | the next-position distribution over word types at each scored word |
+| `simadj_esim/` | per-unit expected-similarity arrays behind the similarity-adjusted cells |
+| `fwd_word_lookahead_entropy_samples/` | look-ahead entropies at depths `1..K` and the per-sample step log probabilities |
+| `manifests/<tag>.json` | the compute driver's provenance sidecar of every exported cell |
+| `verification/<wave>.json` | per-table verification records of one recompute wave |
+| `owt/`, `modelblocks_kenlm_models/`, `kenlm_o1/` | frequency tables the unigram cells read |
+| `pool/`, `pool_s900000/`, `pool_tokens/`, `pool_tokens_s900000/` | the marginal sample pools earlier releases published |
+
+## Layers
+
+Layer 1 holds the predictors REPLICA computes under its own conventions. Layer 2 holds a released defect reconstructed for one study; the `study` column of the index names it. Cells published before the layer field existed carry the layer of their method.
+
+| `layer` | cells |
+|---|---|
+| 1 | 2507 |
+| 2 | 63 |
+| null | 26 |
 
-```python
-import pandas as pd
-idx = pd.read_parquet("hf://datasets/replicaverse/replica-predictors/tags.parquet")
+| `method` | `study` | cells |
+|---|---|---|
+| `continuation_entropy_buggy` | `giulianelli2024generalized` | 1 |
+| `modelblocks_surprisal` | `clark2025wordentropy` | 28 |
+| `wordsprobability_surprisal_buggy` | `pimentel2024probability` | 34 |
 
-# Every word-level Renyi-1/2 entropy cell computed with 512 samples.
-idx[(idx.method == "mc_word_entropy") & (idx.alpha == 0.5) & (idx.n_samples == 512)]
+## Storage classes and units
 
-# What is available for one corpus?
-idx[idx.dataset == "provo"][["model", "method", "predictor", "value_unit", "n_rows"]]
+A `positional` estimator conditions on a prefix ending at a token position and returns a value for that position; its cell is stored per token under `tokens/` and any coarser unit derives from the token table. A `unit_bound` estimator takes the unit as an input, so its cell is stored at the unit it was computed for and `unit_rule` names the boundary rule the estimator used (`whitespace_continuation` is the model's own notion of a word: a token sequence that continues until the next whitespace-initial token). `unit` is the grain of the rows: `words`, `tokens` or `characters`. A carried cell whose method the compute driver no longer accepts has null storage columns.
 
-# Which file holds the raw sample arrays behind a scalar cell?
-idx[idx.tag == "gpt2_mcword_entropy_N512_M20_a1_s0_provo"][["predictor", "file"]]
-```
+| `storage_class` | `unit` | cells |
+|---|---|---|
+| null | words | 26 |
+| positional | characters | 15 |
+| positional | tokens | 429 |
+| positional | words | 1253 |
+| unit_bound | words | 873 |
 
-| column | meaning |
+| `unit_rule` | cells |
 |---|---|
-| `tag` | the cell id, as it appears in the `tag` column of the data files |
-| `file` | the file in this repo that holds these rows |
-| `dataset` | corpus |
-| `model` | the model or frequency table the cell was computed from |
-| `model_kind` | `lm` or `frequency_table` |
-| `method` | how the value was computed (see [Methods](#methods)) |
-| `predictor` | the value key |
-| `value_unit` | `bits`, `nats`, `dimensionless`, `identifier` or `text` |
-| `n_stimuli` | stimuli covered |
-| `n_rows` | rows, which is also the number of scored units |
-| `alpha`, `n_samples`, `max_tokens`, `beam_width`, `max_words`, `seed` | the cell's configuration, `null` where the method has no such setting |
-| `provenance` | where the columns below come from: `manifest` (the runner's sidecar, written for cells computed since 2026-09-06), `unit_metadata` (the Monte Carlo families stamp device, torch build and chunk size on every unit), `records` (the frozen compute environment, the job log that wrote the cell, the scheduler's GPU record, the pickle's write time and the model cache), or `none` |
-| `chunk_size`, `device`, `cuda`, `torch`, `transformers`, `python`, `replica` | the realization: how many samples were drawn at once, the GPU the model ran on, and the package versions. A sample is only reproducible on the same device and torch build |
-| `code_commit`, `code_dirty`, `source_sha256` | the commit of the source tree the cell was computed at, whether that tree had uncommitted changes, and a content hash of `src/replica` that identifies the code even without a commit |
-| `model_revision`, `computed_at` | the Hugging Face revision of the model snapshot and when the cell was written |
-
-Cells published before 2026-09-08 predate the manifest. The environment columns
-are filled only for the Monte Carlo families, which carried their device and torch
-build on every unit. Their `code_commit` was established afterwards from the
-schedulers' job accounting, the checkouts' reflogs, and a file-by-file match of
-the compute tree against the repository; the 50 cells whose tree was overwritten
-before it could be matched stay null, as does every other pre-manifest cell. The
-`history.parquet` file records, for every cell, the commit that first published
-it and the commit that last changed it: the publishing commit is not the
-computing commit, but it is the one record every cell has.
-
-## Row schema
-
-The per-corpus files are long format: one row per (`tag`, `predictor`,
-`stimulus_id`, `unit_index`).
+| `spacy_sentence` | 26 |
+| `whitespace_continuation` | 395 |
 
-| column | type | meaning |
+## Word values derived from token tables
+
+410 word cells are aggregated from a token table of the same model and corpus; `derivation_source` names that table's tag and the cell's manifest records the aggregation. A word takes the tokens whose spans nest inside its span; a token that would have to be split, dropped or interpolated stops the derivation. Log probabilities sum over the word's tokens and the summed value is also stored as `surprisal` in bits; an entropy or distance takes the value at the word's first token. A corpus without word spacing derives word surprisal from the `char_beam` character table instead.
+
+| `method` | rule | derived cells |
 |---|---|---|
-| `dataset` | string | corpus name, which is also the file name of the per-corpus tables |
-| `tag` | string | cell id: model, method and configuration, e.g. `gpt2_surprisal_provo` |
-| `predictor` | string | which quantity the row carries, e.g. `token_entropy` |
-| `stimulus_id` | string | stimulus (passage or sentence) within the corpus |
-| `unit_index` | int32 | 0-based position of the text unit within the stimulus |
-| `text_unit` | string | the text unit itself: the word, or the character for the `meco_char_*` corpora |
-| `value` | double | the number, `NaN` where undefined |
-| `value_unit` | string | measurement unit of `value` |
-
-`text_unit` is the surface string the predictor was computed on, so a row can
-be read on its own and a join against the corpus is only needed to add reading
-times. It is `null` wherever that string cannot be established, which is a small
-and enumerated set of stimuli (see [Data notes](#data-notes)); a non-null
-`text_unit` is the string the value was computed on, never a guess.
-
-## Predictors and units
-
-Log bases are mixed on purpose: `log_p_observed` is a signed natural
-log-probability, kept in the base it is computed in, and every entropy is in
-bits. `value_unit` is what disambiguates them, and the same information is
-attached to the array columns as Parquet field metadata.
-
-<!-- TABLE:predictors -->
-<details>
-<summary>All indexed predictor keys and units</summary>
-
-| `predictor` | `value_unit` | cells | rows |
-|---|---|---|---|
-| `continuation_entropy` | bits | 170 | 3,211,755 |
-| `continuation_entropy_buggy` | bits | 1 | 1,931 |
-| `joint_word_entropy` | bits | 22 | 700,591 |
-| `mc_word_entropy` | bits | 312 | 7,057,470 |
-| `next_word_entropy` | bits | 126 | 1,448,928 |
-| `next_word_renyi_entropy_alpha0.5` | bits | 101 | 1,430,124 |
-| `token_entropy` | bits | 86 | 1,371,270 |
-| `token_renyi_entropy_alpha0.5` | bits | 87 | 1,403,256 |
-| `token_renyi_entropy_alpha2.0` | bits | 83 | 1,365,466 |
-| `unigram_surprisal` | bits | 154 | 2,342,173 |
-| `decontextualised_distance` | dimensionless | 58 | 1,231,728 |
-| `expected_decontextualised_distance` | dimensionless | 58 | 1,231,728 |
-| `expected_similarity_tokens` | dimensionless | 25 | 606,073 |
-| `information_value_context_informativeness_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_context_informativeness_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_context_informativeness_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_context_informativeness_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_context_informativeness_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_context_informativeness_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_context_informativeness_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_context_informativeness_euclidean` | dimensionless | 54 | 639,428 |
-| `information_value_deviation_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_deviation_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_deviation_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_deviation_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_deviation_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_deviation_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_deviation_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_deviation_euclidean` | dimensionless | 54 | 639,428 |
-| `information_value_expected_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_expected_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_expected_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_expected_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_expected_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_expected_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_expected_context_informativeness_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_expected_context_informativeness_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_expected_context_informativeness_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_expected_context_informativeness_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_expected_context_informativeness_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_expected_context_informativeness_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_expected_context_informativeness_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_expected_context_informativeness_euclidean` | dimensionless | 54 | 639,428 |
-| `information_value_expected_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_expected_euclidean` | dimensionless | 54 | 639,428 |
-| `information_value_max_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_max_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_max_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_max_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_max_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_max_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_max_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_max_euclidean` | dimensionless | 54 | 639,428 |
-| `information_value_mean_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_mean_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_mean_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_mean_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_mean_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_mean_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_mean_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_mean_euclidean` | dimensionless | 54 | 639,428 |
-| `information_value_min_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_min_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_min_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_min_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_min_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_min_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_min_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_min_euclidean` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_deviation_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_deviation_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_deviation_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_deviation_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_deviation_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_deviation_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_deviation_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_deviation_euclidean` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_expected_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_expected_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_expected_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_expected_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_expected_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_expected_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_expected_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_expected_euclidean` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_max_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_max_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_max_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_max_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_max_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_max_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_max_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_max_euclidean` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_mean_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_mean_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_mean_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_mean_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_mean_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_mean_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_mean_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_mean_euclidean` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_min_1gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_min_1gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_min_2gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_min_2gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_min_3gram` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_min_3gram_pos` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_min_cosine` | dimensionless | 54 | 639,428 |
-| `information_value_ooc_min_euclidean` | dimensionless | 54 | 639,428 |
-| `infovalue_layer0` | dimensionless | 41 | 665,808 |
-| `infovalue_layer1` | dimensionless | 41 | 665,808 |
-| `infovalue_layer10` | dimensionless | 41 | 665,808 |
-| `infovalue_layer11` | dimensionless | 41 | 665,808 |
-| `infovalue_layer12` | dimensionless | 41 | 665,808 |
-| `infovalue_layer13` | dimensionless | 41 | 665,808 |
-| `infovalue_layer14` | dimensionless | 41 | 665,808 |
-| `infovalue_layer15` | dimensionless | 41 | 665,808 |
-| `infovalue_layer16` | dimensionless | 41 | 665,808 |
-| `infovalue_layer17` | dimensionless | 41 | 665,808 |
-| `infovalue_layer18` | dimensionless | 41 | 665,808 |
-| `infovalue_layer19` | dimensionless | 41 | 665,808 |
-| `infovalue_layer2` | dimensionless | 41 | 665,808 |
-| `infovalue_layer20` | dimensionless | 41 | 665,808 |
-| `infovalue_layer21` | dimensionless | 41 | 665,808 |
-| `infovalue_layer22` | dimensionless | 41 | 665,808 |
-| `infovalue_layer23` | dimensionless | 41 | 665,808 |
-| `infovalue_layer24` | dimensionless | 41 | 665,808 |
-| `infovalue_layer3` | dimensionless | 41 | 665,808 |
-| `infovalue_layer4` | dimensionless | 41 | 665,808 |
-| `infovalue_layer5` | dimensionless | 41 | 665,808 |
-| `infovalue_layer6` | dimensionless | 41 | 665,808 |
-| `infovalue_layer7` | dimensionless | 41 | 665,808 |
-| `infovalue_layer8` | dimensionless | 41 | 665,808 |
-| `infovalue_layer9` | dimensionless | 41 | 665,808 |
-| `next_word_information_value` | dimensionless | 25 | 606,073 |
-| `next_word_information_value_first` | dimensionless | 25 | 606,073 |
-| `next_word_information_value_orth` | dimensionless | 6 | 28,953 |
-| `next_word_information_value_pos` | dimensionless | 6 | 28,953 |
-| `log_p_layers` | nats | 204 | 6,492,037 |
-| `log_p_observed` | nats | 735 | 13,695,105 |
-| `log_p_samples_nats` | nats | 482 | 10,269,225 |
-| `next_log_probs` | nats | 21 | 62,355 |
-| `similarity_adjusted_surprisal` | nats | 25 | 606,073 |
-| `similarity_adjusted_surprisal_first` | nats | 25 | 606,073 |
-| `similarity_adjusted_surprisal_orth` | nats | 6 | 28,953 |
-| `similarity_adjusted_surprisal_pos` | nats | 6 | 28,953 |
-| `information_value_context_informativeness_length` | tokens | 54 | 639,428 |
-| `information_value_deviation_length` | tokens | 54 | 639,428 |
-| `information_value_expected_context_informativeness_length` | tokens | 54 | 639,428 |
-| `information_value_expected_length` | tokens | 54 | 639,428 |
-| `information_value_max_length` | tokens | 54 | 639,428 |
-| `information_value_mean_length` | tokens | 54 | 639,428 |
-| `information_value_min_length` | tokens | 54 | 639,428 |
-| `information_value_ooc_deviation_length` | tokens | 54 | 639,428 |
-| `information_value_ooc_expected_length` | tokens | 54 | 639,428 |
-| `information_value_ooc_max_length` | tokens | 54 | 639,428 |
-| `information_value_ooc_mean_length` | tokens | 54 | 639,428 |
-| `information_value_ooc_min_length` | tokens | 54 | 639,428 |
-
-</details>
-<!-- /TABLE:predictors -->
-
-| `value_unit` | means |
-|---|---|
-| `nats` | log base e. `log_p_observed` is a log-probability, so it is `<= 0`; negate and divide by `ln 2` for surprisal in bits |
-| `bits` | log base 2, `>= 0` |
-| `dimensionless` | a cosine distance or a probability-like ratio, no log base |
-| `identifier` | not a measurement: a hash used to group samples |
-| `text` | not a measurement: a surface string |
+| `char_beam` | sum | 2 |
+| `continuation_entropy` | first | 77 |
+| `first_token_entropy` | first | 40 |
+| `logit_lens_surprisal` | sum | 88 |
+| `responsivity` | first | 36 |
+| `similarity_surprisal_contextual` | sum | 2 |
+| `similarity_surprisal_noncontextual` | sum | 10 |
+| `surprisal` | sum | 60 |
+| `token_renyi_entropy` | first | 95 |
 
-## Tag grammar
+## Row schemas
 
-```
-<model>_<method><config>_<dataset>
-   |        |        |        |
-   |        |        |        `- corpus, also the file name
-   |        |        `- configuration tokens, method dependent
-   |        `- how the value was computed
-   `- HuggingFace model id with "/" written as "_", or a frequency table name
-```
+Word and character tables, and the `responsivity/` files:
 
-`gpt2_mcword_entropy_N512_M20_a1_s0_provo` is `gpt2`, word entropy by Monte
-Carlo, 512 samples, at most 20 subword steps per sampled word, Renyi order 1
-(Shannon), seed 0, on `provo`. The pieces are already split out in
-`tags.parquet`, so nothing needs to parse this string.
+| column | type | meaning |
+|---|---|---|
+| `dataset` | string | corpus, also the file name of the per-corpus tables |
+| `tag` | string | cell id: model, method and configuration |
+| `predictor` | string | value key, e.g. `surprisal` |
+| `stimulus_id` | string | stimulus within the corpus |
+| `unit_index` | int32 | 0-based position of the unit within the stimulus |
+| `text_unit` | string | the unit as the corpus loader spells it: the word, or the character; null when the stored unit count of the stimulus differs from the corpus |
+| `value` | double | the number; `NaN` where undefined |
+| `value_unit` | string | `nats`, `bits`, `dimensionless` or `tokens` |
+| `layer` | int32 | 1 or 2, see Layers |
+| `storage_class` | string | `positional` or `unit_bound`, see Storage classes |
+| `unit_rule` | string | boundary rule of a unit-bound estimator; null for a positional one |
+
+Token tables (`tokens/<dataset>.parquet`):
 
-| token | column | meaning |
+| column | type | meaning |
 |---|---|---|
-| `N<n>` | `n_samples` | Monte-Carlo samples drawn per unit |
-| `L<n>` | `max_tokens` | continuation length, in subword tokens |
-| `M<n>` | `max_tokens` | cap on subword steps taken while extending one sampled word |
-| `K<n>` | `beam_width` / `max_words` | beam width for `charbeam`, look-ahead depth in words for `fwd_word_lookahead` |
-| `a<x>` | `alpha` | Renyi order. `a1` is the Shannon limit |
-| `s<n>` | `seed` | random seed |
-
-Some `fwd_word_lookahead_entropy_samples/` files end in `.shard<start>_<end>`: one
-slice of a corpus that was scored in pieces, covering stimuli `start` to `end`.
-The part before it is the cell id, and `tags.parquet` groups the slices under
-that id while its `file` column keeps them apart.
-
-## Methods
-
-<!-- TABLE:methods -->
-| `method` | predictors it carries | settings used | cells | corpora | models |
-|---|---|---|---|---|---|
-| `char_beam` | `log_p_observed` | `beam_width`=5 | 43 | 28 | 2 |
-| `continuation_entropy` | `continuation_entropy`, `log_p_samples_nats` | `n_samples`=64, 512; `max_tokens`=5, 10, 15; `seed`=0 | 170 | 57 | 2 |
-| `continuation_entropy_buggy` | `continuation_entropy_buggy` | `n_samples`=512; `max_tokens`=5; `seed`=0 | 1 | 1 | 1 |
-| `first_token_entropy` | `next_log_probs`, `token_entropy` | none | 86 | 58 | 2 |
-| `forward_word_lookahead_entropy` | `joint_word_entropy` | `n_samples`=512; `max_tokens`=30; `max_words`=5; `seed`=0 | 22 | 18 | 2 |
-| `information_value` | 108 keys; see predictor table or `tags.parquet` | `n_samples`=100; `max_tokens`=60; `seed`=0 | 54 | 27 | 2 |
-| `logit_lens_surprisal` | `log_p_layers` | none | 188 | 16 | 17 |
-| `mc_word_entropy` | `log_p_samples_nats`, `mc_word_entropy` | `alpha`=0.5, 1; `n_samples`=64, 512; `max_tokens`=20; `seed`=0 | 312 | 56 | 2 |
-| `mmc_unigram_surprisal` | `log_p_observed`, `unigram_surprisal` | none | 55 | 36 | 2 |
-| `mmc_unigram_surprisal_pool` | `log_p_observed`, `unigram_surprisal` | `n_samples`=4e+07; `seed`=900000 | 13 | 13 | 1 |
-| `mmc_unigram_surprisal_pool_ee` | `log_p_observed`, `unigram_surprisal` | `n_samples`=4e+07; `seed`=900000 | 10 | 10 | 1 |
-| `modelblocks_surprisal` | `log_p_observed` | none | 28 | 28 | 1 |
-| `next_word_entropy` | `next_word_entropy` | none | 126 | 58 | 2 |
-| `next_word_renyi_entropy` | `next_word_renyi_entropy_alpha0.5` | `alpha`=0.5 | 101 | 58 | 2 |
-| `responsivity` | `decontextualised_distance`, `expected_decontextualised_distance` | `n_samples`=512; `seed`=0 | 58 | 29 | 2 |
-| `similarity_surprisal_contextual` | `expected_similarity_tokens`, `next_word_information_value`, `next_word_information_value_first`, `similarity_adjusted_surprisal`, `similarity_adjusted_surprisal_first` | `n_samples`=50; `seed`=0 | 6 | 6 | 1 |
-| `similarity_surprisal_noncontextual` | `expected_similarity_tokens`, `next_word_information_value`, `next_word_information_value_first`, `similarity_adjusted_surprisal`, `similarity_adjusted_surprisal_first` | none | 19 | 19 | 1 |
-| `similarity_surprisal_word` | `next_word_information_value_orth`, `next_word_information_value_pos`, `similarity_adjusted_surprisal_orth`, `similarity_adjusted_surprisal_pos` | `n_samples`=50; `max_tokens`=10; `seed`=0 | 6 | 6 | 1 |
-| `surprisal` | `log_p_observed` | none | 442 | 57 | 15 |
-| `token_renyi_entropy` | `token_renyi_entropy_alpha0.5`, `token_renyi_entropy_alpha2.0` | `alpha`=0.5, 2 | 170 | 58 | 2 |
-| `tuned_lens_surprisal` | `log_p_layers` | none | 16 | 4 | 4 |
-| `unigram_surprisal` | `log_p_observed`, `unigram_surprisal` | none | 76 | 56 | 5 |
-| `word_information_value` | 25 keys; see predictor table or `tags.parquet` | `n_samples`=5; `max_tokens`=3; `seed`=0 | 41 | 41 | 1 |
-| `wordsprobability_surprisal` | `log_p_observed` | none | 34 | 25 | 2 |
-| `wordsprobability_surprisal_buggy` | `log_p_observed` | none | 34 | 25 | 2 |
-<!-- /TABLE:methods -->
-
-What each one computes for a word `w` with left context `c`, and the piece of
-the tag it appears as:
-
-| `method` | tag piece | definition |
+| `dataset`, `tag`, `predictor`, `stimulus_id`, `value`, `value_unit` | | as above |
+| `token_index` | int32 | 0-based position of the token within the stimulus |
+| `span_start`, `span_end` | int32 | character offsets of the token in the stimulus text |
+| `token_id` | int64 | vocabulary id under the cell's tokenizer |
+| `token` | string | the token string as the tokenizer spells it |
+
+Array files carry `dataset`, `tag`, `stimulus_id`, `unit_index`, one or more
+list columns whose unit is in the column's Parquet field metadata, and
+`value_unit`. A word-grain file adds `text_unit`; a token-grain file adds the
+token identity columns instead.
+
+Index (`tags.parquet`) columns beyond the row schema: `file` (where the rows
+live), `model`, `model_kind` (`lm` or `frequency_table`), `method`,
+`n_stimuli`, `n_rows`, the configuration (`alpha`, `n_samples`,
+`max_tokens`, `beam_width`, `max_words`, `seed`), the realization
+(`provenance`, `chunk_size`, `device`, `cuda`, `torch`, `transformers`,
+`python`, `replica`, `code_commit`, `code_dirty`, `source_sha256`,
+`model_revision`, `computed_at`), and the storage columns `layer`, `study`,
+`storage_class`, `unit`, `unit_rule`, `context_scope`, `derivation_source`,
+`wave`, `verification`, `verification_file`.
+
+## Verification
+
+Each recompute wave was checked table by table against the corpus loader's unit grid and, where a released or independently recomputed value exists, against that value. The records live in `verification/<wave>.json`; the index carries each cell's verdict in `verification` and the file in `verification_file`. `pass` means no flag was raised; `check` means at least one flag names something to look at, listed in the record's `flags`. A flag that begins `manifest` describes the sidecar the producing commit wrote, so it names a field that commit predates; the index columns are filled from the tag and the method and stay complete. A cell without a record has a null verdict.
+
+| file | verdict | cells |
 |---|---|---|
-| `surprisal` | `surprisal` | `log P(w \| c)` from a causal LM, summed over the word's subword tokens |
-| `wordsprobability_surprisal` | `wordsprobability_surprisal` | word log-probability with the word-boundary and sentence-start correction of the upstream `wordsprobability` package |
-| `wordsprobability_surprisal_buggy` | `wordsprobability_surprisal_buggy` | the same package's uncorrected column: a plain sum of subword log-probabilities, equal to `surprisal` aggregated to whitespace words |
-| `char_beam` | `charbeam_surprisal_K<k>` | character-level log-probability, marginalising the top-`K` tokenisations of the string with a byte-level beam search (`genlm-bytes`), then aggregated to the corpus's units |
-| `first_token_entropy` | `ft_entropy` | Shannon entropy of the next-token distribution over the full subword vocabulary, read at the word's first subword |
-| `token_renyi_entropy` | `token_renyi_entropy_a<x>` | Renyi-`alpha` entropy of that same first-subword distribution |
-| `next_word_entropy` | `next_word_entropy` | Shannon entropy after restricting the next-token distribution to word-initial tokens and renormalising |
-| `next_word_renyi_entropy` | `next_word_renyi_entropy_a<x>` | Renyi-`alpha` entropy of that restricted distribution |
-| `mc_word_entropy` | `mcword_entropy_N<n>_M<m>_a<x>_s<s>` | word entropy by Monte Carlo: `N` samples, each continued subword by subword until a word boundary is drawn (at most `M` steps), aggregated at Renyi-`alpha` |
-| `continuation_entropy` | `cont_entropy_N<n>_L<l>_s<s>` | sequence entropy by Monte Carlo: `N` sampled continuations of up to `L` tokens each |
-| `forward_word_lookahead_entropy` | `fwd_word_lookahead_entropy_N<n>_M<m>_K<k>_s<s>` | look-ahead entropy over whole words at every depth `1..K`, in joint, conditional and marginal form |
-| `unigram_surprisal` | `unigram_surprisal` | context-free word probability read off a frequency table |
-| `mmc_unigram_surprisal` | `mmc_unigram_surprisal` | context-free word probability recovered from an LM by chain-rule Rao-Blackwellisation over sampled contexts |
-| `responsivity` | `responsivity_N<n>_s<s>` | mean cosine distance in the model's input-embedding space from the observed word to `N` sampled alternatives, and among those alternatives |
-
-## Models and frequency tables
-
-<!-- TABLE:models -->
-| `model` | `model_kind` | methods | cells |
-|---|---|---|---|
-| `owt` | frequency_table | 1 | 5 |
-| `wordfreq` | frequency_table | 1 | 56 |
-| `wordfreq-da` | frequency_table | 1 | 1 |
-| `wordfreq-en` | frequency_table | 1 | 1 |
-| `EleutherAI/pythia-1.4b-deduped` | lm | 2 | 43 |
-| `EleutherAI/pythia-12b-deduped` | lm | 1 | 4 |
-| `EleutherAI/pythia-14m` | lm | 2 | 39 |
-| `EleutherAI/pythia-160m-deduped` | lm | 2 | 43 |
-| `EleutherAI/pythia-1b-deduped` | lm | 1 | 16 |
-| `EleutherAI/pythia-2.8b-deduped` | lm | 1 | 4 |
-| `EleutherAI/pythia-31m` | lm | 2 | 39 |
-| `EleutherAI/pythia-410m-deduped` | lm | 2 | 43 |
-| `EleutherAI/pythia-6.9b-deduped` | lm | 1 | 4 |
-| `EleutherAI/pythia-70m-deduped` | lm | 2 | 43 |
-| `Qwen/Qwen3-0.6B` | lm | 1 | 29 |
-| `Qwen/Qwen3-1.7B` | lm | 1 | 29 |
-| `Qwen/Qwen3-4B` | lm | 1 | 29 |
-| `ai-forever/mGPT` | lm | 17 | 897 |
-| `facebook/opt-1.3b` | lm | 1 | 4 |
-| `facebook/opt-125m` | lm | 3 | 35 |
-| `facebook/opt-6.7b` | lm | 1 | 4 |
-| `gpt2` | lm | 21 | 623 |
-| `gpt2-large` | lm | 3 | 34 |
-| `gpt2-medium` | lm | 2 | 26 |
-| `gpt2-xl` | lm | 3 | 47 |
-| `wt103` | lm | 1 | 13 |
-<!-- /TABLE:models -->
-
-`wordfreq` resolves its language from the corpus; `wordfreq-en` and
-`wordfreq-da` pin it. `owt` is a whitespace-token count table over
-OpenWebText, published here as `owt/unigram_counts.tsv`.
-
-Values from different models are on different scales and different
-vocabularies. Compare within a model, not across models.
-
-## Corpora
-
-<!-- TABLE:corpora -->
-**EEG (N400)** (8): `federmeier2007_n400`, `hubbard2019_n400`, `michaelov_n400`, `szewczyk2022_n400`, `szewczyk_federmeier2022_n400`, `szewczyk_n400`, `wlotko_federmeier2012_n400`, `zuco_n400`
-
-**eye tracking** (45): `bsc`, `celer`, `copco`, `devarda2023`, `dundee`, `emtec`, `geco`, `meco_de`, `meco_du`, `meco_ee`, `meco_en`, `meco_fi`, `meco_gr`, `meco_he`, `meco_it`, `meco_ko`, `meco_l1_w2_ba`, `meco_l1_w2_bp`, `meco_l1_w2_ch_s`, `meco_l1_w2_ch_t`, `meco_l1_w2_da`, `meco_l1_w2_en_uk`, `meco_l1_w2_ge_po`, `meco_l1_w2_ge_zu`, `meco_l1_w2_hi_iiith`, `meco_l1_w2_hi_iitk`, `meco_l1_w2_ic`, `meco_l1_w2_no`, `meco_l1_w2_ru_mo`, `meco_l1_w2_se`, `meco_l1_w2_sp_ch`, `meco_l1_w2_tr`, `meco_l2_w1`, `meco_l2_w2`, `meco_no`, `meco_ru`, `meco_sp`, `meco_tr`, `onestop`, `potec`, `provo`, `sbsat`, `ucl_et`, `zuco1`, `zuco2`
-
-**maze** (1): `natural_stories_maze`
-
-**self-paced reading** (4): `brothers_kuperberg`, `brown_spr`, `natural_stories`, `ucl_spr`
-<!-- /TABLE:corpora -->
-
-## Files
-
-<!-- TABLE:files -->
-| path | files | cells | rows |
-|---|---|---|---|
-| `(root)` | 58 | 1827 | 121,571,006 |
-| `cont_entropy_samples/` | 170 | 170 | 3,211,755 |
-| `fwd_word_lookahead_entropy_samples/` | 22 | 22 | 700,591 |
-| `lens_layers/` | 204 | 204 | 6,492,037 |
-| `mcword_samples/` | 312 | 312 | 7,057,470 |
-| `next_log_probs/` | 21 | 21 | 62,355 |
-| `responsivity/` | 58 | 58 | 2,463,456 |
-| `simadj_esim/` | 25 | 25 | 606,073 |
-<!-- /TABLE:files -->
+| `verification/wave1-tokens.json` | check | 406 |
+| `verification/wave1-tokens.json` | pass | 6 |
+| `verification/wave2.json` | check | 269 |
+| `verification/wave2.json` | pass | 626 |
 
-| path | contents |
+| flag | tables |
 |---|---|
-| `tags.parquet` | the cell index described above |
-| `history.parquet` | one row per file ever published: the commit and time that first added it and that last changed it, and whether it is still present |
-| `manifests/<tag>.json` | the runner's full provenance sidecar for cells computed since 2026-09-06: resolved options, model snapshot, corpus hash, source-tree hash, package versions, and the environment (commit, device, CUDA build, time) |
-| `<dataset>.parquet` | scalar per-word predictors for one corpus, one file per corpus |
-| `responsivity/` | embedding-distance predictors, same long schema, one file per (model, corpus). Has its own README |
-| `next_log_probs/` | the model's predictive distribution over word types at each scored position |
-| `mcword_samples/` | the per-word Monte-Carlo sample pool behind `mc_word_entropy` |
-| `cont_entropy_samples/` | the per-word sample pool behind `continuation_entropy` |
-| `fwd_word_lookahead_entropy_samples/` | word-level look-ahead entropies at depths `1..K`, plus the raw per-sample step log-probabilities |
-| `owt/unigram_counts.tsv` | OpenWebText unigram count table |
-| `modelblocks_kenlm_models/` | order-1 KenLM ARPA trained on OpenWebText |
-| `kenlm_o1/` | per-word OpenWebText unigram surprisal, one pickle per corpus |
-
-Only `tags.parquet` and the per-corpus files are dataset-viewer configs. A row
-holding a vocabulary-sized array does not render usefully, so the array folders
-are loaded by path.
+| manifest identity lacks storage_class | 405 |
+| manifest layer is None | 405 |
+| words empty or containing whitespace | 274 |
+| context scope story (layer 1 conditions on the whole stimulus) | 49 |
+| BOS disabled | 42 |
+| context scope story | 28 |
+| context scope sentence (layer 1 conditions on the whole stimulus) | 14 |
+| unit characters | 9 |
+| manifest lacks the window rule | 8 |
+| manifest sampling block does not record positions: all | 8 |
+| manifest sampling block records no seed | 8 |
+| context scope sentence | 7 |
+| token spans differ from the independent tokenization | 7 |
+| NaN beyond the first token | 7 |
+| NaN beyond the first word | 4 |
+| unexpected producer commit | 2 |
+| non-finite values | 2 |
+| independent per-token recomputation differs by 0.0017 | 1 |
+| stimuli whose stored words differ from re-aggregating gpt2_simadj_surprisal_noncontextual_tokens_sbsat.pkl | 1 |
+| stimuli whose stored words differ from re-aggregating gpt2_simadj_surprisal_noncontextual_tokens_zuco1.pkl | 1 |
+
+## Coverage and gaps
+
+`coverage.parquet` has one row per planned (method, model, corpus, unit) table and one per produced table no plan names. `status` is `exported` when the table comes from the latest release, `carried` when the store ships it from an earlier one, `quarantined` when the compute driver refused to publish it (a stimulus failed or the values failed an audit) and the store has no earlier table, and `absent` when the store has none. `reason_class` is the plan's class for the table, one of `complete`, `excluded_by_decision`, `expected_refusal`, `held_decision`, `known_gap_beam_empty`, `missing`, `running`. `expected_refusal` is a whitespace-word estimator on a corpus whose words are not the model's words, and `known_gap_beam_empty` is a character beam that emptied on a stimulus. `detail` holds the failure log of a quarantined table or the command of an absent one.
+
+| `status` | tables |
+|---|---|
+| absent | 1 |
+| carried | 43 |
+| exported | 1309 |
+| quarantined | 5 |
 
-## Loading
+| `status` | `reason_class` | tables |
+|---|---|---|
+| absent | held_decision | 1 |
+| carried | excluded_by_decision | 1 |
+| carried | expected_refusal | 6 |
+| carried | held_decision | 24 |
+| carried | known_gap_beam_empty | 2 |
+| carried | missing | 8 |
+| carried | running | 2 |
+| exported | complete | 855 |
+| exported | expected_refusal | 2 |
+| exported | not planned | 452 |
+| quarantined | held_decision | 5 |
+
+| quarantined `method` | tables |
+|---|---|
+| `modelblocks_surprisal` | 1 |
+| `wordsprobability_surprisal` | 2 |
+| `wordsprobability_surprisal_buggy` | 2 |
 
-One corpus, one cell:
+## What changed in this release
 
-```python
-import pandas as pd
+Built on the published store at commit `2508a73d9431d94a3b776ced13c3fc44f49d9427`; every published cell this release does not replace is carried over unchanged.
 
-REPO = "hf://datasets/replicaverse/replica-predictors"
-provo = pd.read_parquet(f"{REPO}/provo.parquet")
-cell = provo[provo.tag == "gpt2_surprisal_provo"]
-cell[["stimulus_id", "unit_index", "text_unit", "value", "value_unit"]].head()
-```
+| wave | cells exported |
+|---|---|
+| wave1-tokens | 412 |
+| wave2 | 895 |
 
-One predictor per column, ready to model:
+- 825 published cells replaced by a recomputed table of the same tag
+- 0 published files deleted because their cell was recomputed without the array they held
+- 1895 files written
+- new: `tokens/` (token-grain tables), `characters/` (character-grain tables), `coverage.parquet`, `verification/`, and the storage columns of the word tables and the index
+- array keys without a store folder, left in the pickles only: `orth_similarity_samples`, `pos_match_samples`, `sampled_words`, `similarity_samples_tokens`
 
-```python
-wide = (provo[provo.tag.isin(["gpt2_surprisal_provo", "gpt2_ft_entropy_provo"])]
-        .pivot_table(index=["stimulus_id", "unit_index", "text_unit"],
-                     columns="predictor", values="value")
-        .reset_index())
-```
+### Carried cells on an older unit grid
 
-Through the dataset viewer:
+133 cells this release carries were computed against an earlier version of their corpus, so their `stimulus_id` and `unit_index` do not line up with the cells this release stages into the same table. Join them to a recomputed predictor only after checking the units match. `tags.parquet` gives each cell's `n_stimuli` and `n_rows`, and a recomputed cell of the same corpus gives the current shape.
 
-```python
-from datasets import load_dataset
-idx = load_dataset("replicaverse/replica-predictors", "tags", split="train")
-provo = load_dataset("replicaverse/replica-predictors", "provo", split="train")
-```
+| table | cells on an older grid |
+|---|---|
+| `dundee.parquet` | 20 |
+| `meco_de.parquet` | 2 |
+| `meco_ee.parquet` | 7 |
+| `meco_fi.parquet` | 5 |
+| `meco_gr.parquet` | 6 |
+| `meco_it.parquet` | 5 |
+| `meco_l1_w2_ch_s.parquet` | 6 |
+| `meco_l1_w2_ch_t.parquet` | 6 |
+| `meco_l2_w2.parquet` | 10 |
+| `meco_tr.parquet` | 5 |
+| `onestop.parquet` | 17 |
+| `provo.parquet` | 15 |
+| `sbsat.parquet` | 20 |
+| `szewczyk_n400.parquet` | 9 |
 
-Raw distributions and sample pools, by path:
+## Loading
 
 ```python
-import numpy as np, pyarrow.parquet as pq
-
-lp = pq.read_table(f"{REPO}/next_log_probs/gpt2_ft_entropy_provo.parquet")
-print(lp.schema.field("logprobs").metadata)          # {b'unit': b'nats'}
-dist = np.asarray(lp["logprobs"][0].as_py())         # (n_word_types,) log-probs
-np.exp(dist).sum()                                   # 1, to float32 rounding
+import pandas as pd
 
-la = pq.read_table(f"{REPO}/fwd_word_lookahead_entropy_samples/"
-                   "ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_de.parquet")
-row = la.slice(0, 1).to_pylist()[0]
-steps = np.asarray(row["log_p_step_nats"]).reshape(row["n_samples"], row["k_words"])
+REPO = "hf://datasets/replicaverse/replica-predictors"
+idx = pd.read_parquet(f"{REPO}/tags.parquet")
+provo = pd.read_parquet(f"{REPO}/provo.parquet")
+cell = provo[provo.tag == "gpt2_surprisal_provo"]
+tokens = pd.read_parquet(f"{REPO}/tokens/provo.parquet")
 ```
 
-## Data notes
-
-**Text units.** `unit_index` indexes the corpus's own unit layer, which is
-words everywhere except the twelve `meco_char_*` corpora, where it is single
-characters (including spaces). `text_unit` is that layer's surface string, so
-for a multi-word interest area it is the whole area, with internal whitespace
-written as `_`.
-
-**Missing text units.** A stimulus whose stored unit count disagrees with the
-corpus was scored against a different version of that text, so no position in
-it can be named and every one of its rows gets `text_unit = null`. Six
-`brothers_kuperberg` stimuli are in that state: `bk_009_{H,L,M}` and
-`bk_058_{H,L,M}`, in all of that corpus's cells. Their values do not correspond
-to the shipped sentences, so filter them out rather than joining them.
-
-**Stimulus ids.** MECO first-wave rows previously carried a float-formatted id
-(`meco_en_1.0`). They are now written as `meco_en_1`, matching the corpus.
-
-**Distributions.** `next_log_probs/` holds the model's next-position
-distribution after marginalising the subword vocabulary onto unique surface
-forms and renormalising, so a row is one log-probability per word type: 32,761
-entries for `gpt2` and 79,265 for `ai-forever/mGPT`. Each row sums to 1 under
-`exp`, to float32 rounding. It re-derives any distributional statistic over
-word types offline (entropy, Renyi at any order, top-k mass, the observed
-word's surprisal).
-
-**Precision.** Scalar predictors are float64 as computed. The stored
-distributions and sample pools are float32, so a statistic re-derived from them
-is not bit-identical to one computed in float64.
-
-**Mixed-unit files.** A row of `fwd_word_lookahead_entropy_samples/` carries six array
-columns rather than one value, and they are not all in the same unit:
-
-| column | unit | contents |
-|---|---|---|
-| `joint_word_entropy` | bits | H of the next `k` words jointly, for `k = 1..K` |
-| `conditional_word_entropy` | bits | the step from `k-1` to `k`, same length |
-| `marginal_word_entropy` | bits | H of the word at depth `k` alone |
-| `marginal_word_entropy_plugin` | bits | the same without the small-sample correction |
-| `log_p_step_nats` | nats | raw per-sample step log-probabilities, row-major `[n_samples, k_words]` |
-| `id_hash` | identifier | word-id hashes for those samples, same shape |
-
-Its per-row `value_unit` reports `bits`, the unit of the entropy columns, and
-`tags.parquet` names `joint_word_entropy` as the cell's predictor. The exact
-unit of every column is in that column's Parquet field metadata, so read that
-rather than the row when you use one of the other columns.
-
-**Coverage.** Not every (model, method, corpus) combination exists. Check
-`tags.parquet` rather than assuming a cell is present.
+Values from different models are on different scales and vocabularies;
+compare within a model.
diff --git a/celer.parquet b/celer.parquet
index a984b73f7d967721c21523a00d5952d879d5303a..47f7a03f58b6dbd9b0207e9e2c9227c99e1af299 100644
--- a/celer.parquet
+++ b/celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fb00d307a0f0419f6448afb9049bb45fc3c8344be3d5f512381563ab91425b57
-size 183044631
+oid sha256:bdd50cec403dddae96d386f1ffbf4e1131d1235a0308caad6125098656adfb09
+size 175160031
diff --git a/characters/celer.parquet b/characters/celer.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..d7808135fbb1dc0ed1a5e20977d5cc54399b4acd
--- /dev/null
+++ b/characters/celer.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48c5e3809b1128f40fd4b79da06ad102fa5b9018b785939e4e20c4383aeb5f4e
+size 16119649
diff --git a/characters/geco.parquet b/characters/geco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..d8bbe012759189c2a65a0bc0ac9c8cfda6f9be75
--- /dev/null
+++ b/characters/geco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:508a4c983b6630f6604c7c7af35ff87de6f211b19020313b849528d9423bf258
+size 2774048
diff --git a/characters/meco_l1_w2_ch_s.parquet b/characters/meco_l1_w2_ch_s.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..465d798b23e525d6a39246ce1f0f60a7f6ff7be7
--- /dev/null
+++ b/characters/meco_l1_w2_ch_s.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7b3dbe92ccb8b61a43ef6f6e5ac40e8ebc9a76edc76e9c1276c3519f3b67474
+size 47373
diff --git a/characters/meco_l1_w2_ch_t.parquet b/characters/meco_l1_w2_ch_t.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..d5ce34d5c49b9d2011edef068c1e96653e878cc0
--- /dev/null
+++ b/characters/meco_l1_w2_ch_t.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fca32d4e0596a077d930dfae1f819816616606d88621afb2590d39f9a62b3be2
+size 47751
diff --git a/characters/provo.parquet b/characters/provo.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..b0518fd107eeb486d9c09477ee0d464d52187b09
--- /dev/null
+++ b/characters/provo.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13892c844264b0816f23a3fd2ff177668cdbe83cc0470ad7e1deaf953d5dc1e1
+size 317086
diff --git a/characters/sbsat.parquet b/characters/sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..f982da24e44e3ab209fe9ddab25e480a82fb11a7
--- /dev/null
+++ b/characters/sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d0c4920ea596b70fea74ad405ee6e91d6d044fa116bf67cbd3e8b1d50c28696
+size 277078
diff --git a/characters/zuco1.parquet b/characters/zuco1.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..36d3084553e94e6c21b961bdd9d92509577f8dcb
--- /dev/null
+++ b/characters/zuco1.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f59934a58a13ed7148948f45f7063b2d987d895eb070974f9c1ce25ed5be5609
+size 392503
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_copco.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_copco.parquet
index c7138dc464580fcb76b3928f235168bcccd788de..f5b4f7e17a6cc342d648aeaccc7139aa0d72821d 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_copco.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_copco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:15710ac3b13170302bfe912069fb2b99994a1b51b91f9d6449d132c756427cd5
-size 58235397
+oid sha256:851f8feeeb278d74f8e8e870363ae8e5e56e9054286c7db2397f0dd52c3d05ab
+size 58233348
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_geco.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_geco.parquet
index e2cdfb354b991d7d1b0d213398a5813802778ff7..5f13d82678d436f7ab9e42e7f16b30cb1a71d36d 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_geco.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e5167de17a9c2ad04da782050c7a654f99143597d713ecca454d0ae1013ad08a
-size 102039855
+oid sha256:c3fdff73dd3ede0cda3f8679df1ad798a8d622bd2946127dc81b9fc61a7b0556
+size 102037436
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_de.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_de.parquet
index 0b9b5cfd611ce8694b5bcf25ed99edb692b02157..aaf4fdb9a991b3a7fbb9080814095608d610fceb 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_de.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:beeb3c1c5230d37faa961cb34ff2d3a31e755692e3e2e74586ae4f3899a7280a
-size 4266903
+oid sha256:6dc92af684e5078827a0c4d0d7f524a264481d75e57d5d88b32b085858b4077a
+size 4266131
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ee.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ee.parquet
index a7d228b0a5d8fb6b94ac0c5c44fa55483f4a88d0..45cb4f447dbedd8c5f415d4ac826e1f81462e089 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ee.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:915c58037288daca91dfcefe899609e55a72ff4000e4f7e41a25694f8065279c
-size 3355044
+oid sha256:ee9ef4618e7d0b51de08820aed0b023861e052610d4602397b70d541b06e8e20
+size 3351995
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_fi.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_fi.parquet
index 6d9eee5c9109e6e7a430f99c662ea813fe31d3fa..b5b07af8c8e1cb557ac496ecf32b8674f7293b30 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_fi.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:415a7d554d5734dd341a809ca456c5484e167fb8fbb60d5afdc2bd53a4400fe5
-size 3292674
+oid sha256:2aa563395222f6b677783e464f258bc4b669cfbf4f2aff0b6ae9c8ede5c57a96
+size 3290289
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_gr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_gr.parquet
index 27e37ccaaec8622a5c1c927786fd7627c06f9259..8899db9d8e136c10302c52f41468d00dc50bead3 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_gr.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f9ec77b135f9cbf90aa2a2787404140cf9b1de17ab3e1b9d92436777a8865343
-size 4320468
+oid sha256:1751121de40994333b00e3762850715cbd561fbb2de29ee1ade384b3fc215900
+size 4316239
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_it.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_it.parquet
index ed1f783f2867d0bf1a9d10ed3e6713dfe630db7b..7ca1577275581bd1496d8e1efb9529f38a2c2237 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_it.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d00239e39d04e8e471dc64fd47e9c7891fc2dc3d527ada841e8222946cc92a86
-size 4417652
+oid sha256:de280422d9c4b7101a6765b6e9454da14df419ab41efe6e0eb5526f13d15246d
+size 4416321
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ko.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ko.parquet
index c289e1d6b6bcbc3eb8f568220f90d4c632cf03ed..670bf6ac8024068293a40edc43911903820bccbc 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ko.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ko.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b164d46a94cd777662f155e0ae6d511a90d5d3bf59bc091b5422acb98031ad9e
-size 3844341
+oid sha256:198b35c7523cc7b92b578c04fb4a7d520960eb485782fb496808c4beade7835d
+size 3843030
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_tr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_tr.parquet
index 2665d294cf15f7b871cd7da82cf939fdd51275d1..53b78143b2bc3c806a81ac1ececbb6ea166a8717 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_tr.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:28a9c3dc091df458a88cf6fcf5a7cd86010928809eac00146dded5fd1502b68b
-size 3664599
+oid sha256:1f1026e74de714d35daf819483691c9504b1d8f64dd97a0007228db0c1feefc3
+size 3662575
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_onestop.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_onestop.parquet
index af30f268c8d29a8974d43cb21db0e4b8ce2ac17c..11ffb5377856748ac36f366f0d9660399aff70c0 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_onestop.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c554a29bcffeee22cd6b3c3c70a4c174efd338099f9ff5461cb5c6174927bdca
-size 65463417
+oid sha256:3facb0580ad224bb3f579c6e213c8439ed0578389941d6414af97b9f90859580
+size 67206366
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_provo.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_provo.parquet
index b42b9d372dbc36d3f234dff3ac4f12400ee89311..c53da1e0270afd57d3f15887a698570bc97b35b5 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_provo.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ce449bcd68eb95c2ee191123c311a49ea791a91a0a2012b31f01c9442189410e
-size 5456688
+oid sha256:329076a9030c5747521bcdc6d9d65fba92b61314e0d9f7b96e6fbea937afe401
+size 5554960
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_sbsat.parquet
index 6eda778330e84a26627a23226095c245ca52ddb8..82d5c492a9cf68fc108304f260cc4234adf2fa76 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_sbsat.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7381203f33bc836e50eb71b280cb75d8d801deaaa287b61d228ae472b9edfcbd
-size 5064724
+oid sha256:646ddca0ef397af52d029c983ce80a6b2858c9363335502b1712f5cbee95e447
+size 5062645
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_story_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..b3faf3016243e4ece2cb8533b8730d5adeeb3ac0
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07f169e222f33ca1fae8bbd8e3f6c2664eb5e553f5ea11ffc3e2e7b9a31c582c
+size 5063222
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_copco.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_copco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..6987d2ee31a68dfad195f0edc63f245af24821d8
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_copco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a94dfbc6fc3943400323364a759f343bc07e29d3f975d84cd942bad3c8289f8b
+size 100358719
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_geco.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_geco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..744fe4df23ad3c5f6ec0f4e654af213ec535901e
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_geco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6843d42c884822b23a3236eec58d566898054c7851327143044e81adbe487a5a
+size 149141671
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_de.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_de.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..9078212937649b287d76cb70695e91522a5c69d6
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_de.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28df2e44f0af9ce1f35ac717f813c9b9090f6ad8f76be21cdb1d65ae0af040fe
+size 6424230
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ee.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ee.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..bd00f4db418749daa39170c369cd0494e586cbf3
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ee.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e251b7c9d88fa8b6f43e05559d1b7ac1a26b95c15da6721131ca96daae99710c
+size 8973104
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_fi.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_fi.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..20b576c9fec02410ea23e05059c58f4262716e6e
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_fi.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9077305657e6f6af4931346dcb27e882a67cd6f22ed69ed1459791598145106
+size 7671814
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_gr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_gr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..a85ee8129b9743db0bd32bb05f47ec0836e74b28
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_gr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19f2f1e65a245b0830342cfbe8f659b069d228802645594848915eb1fd028bfd
+size 9229370
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_it.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_it.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..c7db7b8055c52ca38cd3269dc850c183f15e0a11
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_it.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2a8c65589f352eace27e982d349a0af009cf3b64e500f98072176c0834c745b
+size 6504167
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ko.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ko.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..25bca793cd3e6561b7794a159e8efdf0078f030a
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ko.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7967bb92f4e70fd947e834a58da856aa38174476156412da8801040651c1e4ca
+size 8808824
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_tr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_tr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..006e937aa74272707e2f0630a017193d1e4c427f
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_tr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89fc78cc99e00e02eaafccd704dbf77ae58d087d721c6e7a27d6577e03ac13a5
+size 7614215
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_onestop.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_onestop.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..8dbabeef1e36c9a0b308340b210006ea26fe6d74
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_onestop.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12b960392f3c9bf97cf48d35204468a43896beeff7d6ca6e15bc35019d9dc907
+size 88208328
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_provo.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_provo.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..aaebfb1002a9cd6ffa670f59c03d88fa82c21df2
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_provo.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cec3742a9ced108403e924864370b8730c7d05f29fe8b7e10026be776e454b28
+size 7013577
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..f80c61469a6dd791a8bda60f7e00326069ff18d7
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e9135359c4c2331948075b222273cd80306fc36d72c61e3de1a27538876483e
+size 6548713
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_story_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..36481522076001fde1de609974ea90bb9fa8d860
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:281fd9dd1f5680edfafd3ab8209db2aed3bf0c679579f7a855d84abb630f3b7e
+size 6536120
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_de.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_de.parquet
index e15278655d765f016bb622879b561b417b7d8dfb..a6e0e8b7554749f3da769e226d1af2e9e96e4d31 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_de.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e3538b20d406e5e9f49dfbb0a685cba9d806fed8c27af10bf0a91cec368e966a
-size 4219091
+oid sha256:fdae73b0fd754edf58874d4db69d13a3872ae32b71b405419b8bf91ab7f02a3f
+size 4221557
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_fi.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_fi.parquet
index 58c16889cd57ddb9296638c8711ca228c4c9f467..031f99050e6a58b9f002998f3ae117ab0c10e5be 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_fi.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:687b87dadecf33f74c048ee448dd1f09829f4094757de3f98bc4ee1a8918c09d
-size 3274190
+oid sha256:0814bb70d3ff20ebdbcd93b53cc14b92ffc3b512d5246a77c5aad1e89b9eaba7
+size 3270206
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_gr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_gr.parquet
index 22c3e84085f742f5b2d9a3748de51445d230f62c..598be98c22f36febfe28777fdf66c57966e807b7 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_gr.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aa98b53c9d5ff58aab2ca99fab212c85c07c22533c96acc984bae6fbf6059d57
-size 4372259
+oid sha256:f522d549564120150e2a0136debb8d3aa65311c47c62e0b7b600482534b7e944
+size 4368349
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_it.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_it.parquet
index fbaabbdb31d4a76439ee792c30bfb4446ea73d16..eb908fa4a857ac03fe6b7edddf7d871a2f7b88e5 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_it.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:028d15f4307f8aaf65f4ef7e33931ac720c5c3c2f516c8e99701d36437682208
-size 4402928
+oid sha256:bd3514dff08c5cc59a3801b49ad7d6a46eaa31b756d4f5bcd99b56c4cc6cdc6d
+size 4401279
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_ko.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_ko.parquet
index 4a80b3b1c566d38f47883c7ec6ae25fd9e1d03f6..96812293dd5e68b6b16e60f41d345a2fb516f2cb 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_ko.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_ko.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9dbed8262e752ec022350cc9af895e007a4a9585284632a8249f8f4a5c27d97b
-size 3832750
+oid sha256:46054cb8017bdfebf9e6aa72b675e6da9c79f6f9375f6ed4b3b7898e698811e8
+size 3830215
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_tr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_tr.parquet
index 60590b87dc9c692ce66cff533c209f7ae25b88f0..3110acb56554fcdb9e965af3dd0d4f593b92e030 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_tr.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b21fd0a3bcaa4e20bc2cca1a34ffc06f3af96af87a1062dfbd601c114f0131fd
-size 3670482
+oid sha256:b5ad3779b7dfa95e0cc383e6bdfb4e5cfe1444323b068e1e478c818ecaeef862
+size 3667847
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_de.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_de.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..9d0e80e09607cc4cd07f3e226d676a779f750c31
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_de.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b1260d1d774b3e850682d19f0ed6269534740622009237ea8c6de28d2e68d98
+size 6363351
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_fi.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_fi.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..b952356b75850d804b4a2830ca77a7237c60ec79
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_fi.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb90577d0914e3c932aff6a3ff7fdd98768ec3fe1012e56ef17f8618d5b6740a
+size 7657217
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_gr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_gr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..37b0403ae5a59a9d04038e2be2c7fa2c7f4524df
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_gr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1441b3e1a0033365e1e8a902293a41ce69e98e19d50afcaeb2a0635125830a7a
+size 9333792
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_it.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_it.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..2b72bf2513ca30ba8e573d5a6ccc27e3e5e73d5f
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_it.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ee4e89834f61a862b62df2c8200774f788e506c37f7a5525c0ebcc23ba08501
+size 6492976
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_ko.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_ko.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..376f8f154dca9414617f9aa4141a97e152eb9dc2
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_ko.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a927c20da3a89d5b431a94b4489e1cc30c69bfe7de19960908d0c591e1dcf81f
+size 8800239
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_tr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_tr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..2bae57778249e465fc9f5ba2ca0250316dfa05d9
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_tr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6752254f2dd756730f555a70d15f50fb74ab0b839709e2a11a5a984651d5733c
+size 7643846
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_celer.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_celer.parquet
index 229fd642c4d55aedf8445c74a4b94877be0ce1b1..208cc2b539934aa09c3d7b01bfeb9d55f9b1c5ad 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_celer.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7763a2823806dbadfc6ed2904a2c8a585403ba2db9fd48421cb93302b74b701e
-size 571826464
+oid sha256:10a3d3cb614e02a1ec899b3b7d5e7b0465f45d626d62abe04ced323788459dad
+size 570354449
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_copco.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_copco.parquet
index 575395ce3c7de78a3d0d04c11b6280ad093d68c5..383876f50d81a686104d0250e63a380282fe61e0 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_copco.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_copco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4c584af97b20fe8b2351b4c67999f1e73b21e4b6d8d4aea3139efa5208cec51e
-size 57754086
+oid sha256:2ca95af52a13a93f79ae17acb40ce9f76e486d87f36118601d952fa6719b1b87
+size 57756731
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_dundee.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_dundee.parquet
index 9cf5ff62855a2aca5c7b698a713b8ae6cdec7655..6d0437655f49fbf3310ee118d04011fb807372cc 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_dundee.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:81a77cab9cf596180b6e4a65c22d871ac03bcb281e8af191677030a41bfad73a
-size 90173070
+oid sha256:2833b927a1ef2da8fbcfa0862b9371e26440317a26705d60ecc7049204ae9267
+size 91721833
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_federmeier2007_n400.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_federmeier2007_n400.parquet
index ddfe481f4a92ae211b312d691c036baba0d28500..0a00504052cd568235d02da00c910f20d16fc2b6 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_federmeier2007_n400.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_federmeier2007_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1af67a03fa99aa587d7aba2c47975e4cfdb7b974eaae4e493d724b9d186fe963
-size 11544036
+oid sha256:fdf6209c38c3e99b4bdf91976dbf58a27a6407a5c0bfc1fa574c16622384a47e
+size 11543655
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_geco.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_geco.parquet
index 82694b357817ed2f82c3b9476294b12477b87b7a..1876eb3b9c2d1c3086be4d7fbddcb6693bd5019a 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_geco.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:17efbd56917fa5d7e57c4935ba43baa0c2aa5d594fb9305b0487b240aab35f60
-size 100413725
+oid sha256:461ea977d5f34259db5ec1284258456a4a8cd8aa3f463669a4c3d7ac8aad91c6
+size 100429010
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_de.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_de.parquet
index 54f13dccb6591ef6e9e9046de1454249f1da6b3d..0a768ab83a8717534fc7353451cbb681962b230c 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_de.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9f366a5cd11e9e4e1a476141b394cae0ddca67b248b0248e139741781a4dfc05
-size 4219034
+oid sha256:0466733456b44741341dcd453947ed920d04a49bb846d7ac901077e8bcf4254b
+size 4214222
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ee.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ee.parquet
index 1fbb683d2d1483ff20ee662cbe866bc1a2643407..d74ed8d5ee16d079a63a4d43dbf712a24d1cb518 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ee.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:89734dc3eccc44d20205134f377400a3e8f6456635533a35f6b81f3b9a41180a
-size 3384231
+oid sha256:afac4843bf9e663fe95cc4e9a826f3aea22faa75b60b413ef50c2bd0591b64b8
+size 3379012
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_fi.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_fi.parquet
index 50d9a7c51ef8e98b4fa70cb729ca11327766da8c..dad5a7d93d47beb81b025b2a86171c877c7918da 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_fi.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:449c6e3e44d6f04d8b8367a22726e35fab92866ab2a9b207013a37b3dde1f8e1
-size 3238904
+oid sha256:79c5a56c0e221d8b03cfcfd3b2e8c931a9888b4d57a89699e116f78e3e9875a2
+size 3231445
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_gr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_gr.parquet
index f1935665d62ac5777a1371e8dc0b797005bc7f4f..20254452f99dfcec2f25b77b05eb59a46768737b 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_gr.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:31bb8c539323da34c2e5ca1d66659ba2893a322e5feed01612cbfbefc3130953
-size 4272397
+oid sha256:ea08c809d8e32d2f2a129b8bf1e1e8609f5a023d992a023da3b7abc066030f93
+size 4263473
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_it.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_it.parquet
index 251f9a5897e7aebec7a0310a086a0f8ede38dfd5..2757a26c03494a91e9fc9e416a531161facad722 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_it.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cf0ee805e179ff6a0f4e7a3376afd835b1aa8149f5017bcc9b922545e7eaa97d
-size 4375359
+oid sha256:9396d9dbf7687d8342fc87b90e32c4d1b453764d5523780cd078c8d017610b19
+size 4377631
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ko.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ko.parquet
index e5f1e079b48255bf051b5d15ef8653ae88d02b53..334d75c3caac380b30a15a0b0f8c7c264ad9da2d 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ko.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ko.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5dda6ad9ab461abe60ba56dd4d2f96b2bdf5fae6b7c5904ac7df641b5bf71c98
-size 3791107
+oid sha256:27e19c88e4b0afc6ea00183ce3e5d16d39f67930293ebbbe739f2f300612588d
+size 3791221
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ba.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ba.parquet
index 229063e39cec5849f4a1c31afab514a298577e67..7ca3337dd58848ce9b54fc1c517b0c2eb27939a3 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ba.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ba.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a255ecea25393f47d3209e3a76e6b9a96bc7d5dc9b2d56203ae10d6c1554f9ed
-size 3630416
+oid sha256:db2d82a38f6bea1d61b388f9917fa38d7398ea57dc8ae102083daa21d6a59db7
+size 3628653
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_bp.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_bp.parquet
index ec25b105b65760ceb1fc83b456d1aaa457e2dcea..d397202cf62c341ac36ce39cd2d4b00b54527ff3 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_bp.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_bp.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9efcf15b77a07e2b57a27ae5cc1ebd500b46c956332d1b0c0d4cde4a45967c70
-size 4536325
+oid sha256:44d990329ff0fa6be0e8cb1407df0b99bc692e6d23080480e51fe0cd2a9ff03e
+size 4530844
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_po.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_po.parquet
index 81be7fd91d18a8103aad532aeb300b4d0aef9a26..4e3c7570ca0cc0919057cb3cfb7d48cc62d55856 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_po.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_po.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1022bf2dbd21fa4172548c4e09c4cd209490c54a07cdeb353035abb33aa27eee
-size 4198852
+oid sha256:96112142e4ac9154e7ed4d3a82778d657b3b572e339eb20ac8b7539f44bfaac3
+size 4199579
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_zu.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_zu.parquet
index a349156edbe3b236534d5ae6e994e353585851ec..b48daba0f6622365052df18c2aa8ec1f19913a16 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_zu.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_zu.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:448086ebeaa90793529b8b27cb45b8411ba7e57cb0ffd17bc0b2021e4595bf0b
-size 4198852
+oid sha256:6b022d4fc23c78f99a1cd5be1bf1481eda343fc5a944ab79134d7a74c8cabc80
+size 4199579
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ic.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ic.parquet
index 01e237e56e9cb0b2a2fa6993f38942aa61c862f1..9b7fae68e63f2b099cefb2cc36d1c79cc1ced3ef 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ic.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ic.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9a35215924060e1c17a40009ca039750754be9af6dbf6ae8165860f209c472a0
-size 4400839
+oid sha256:798cb4ad7005554a01fdb226b6977253442c7efdc281f820dae455a7fb1a401e
+size 4399365
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_tr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_tr.parquet
index f8e06c126d7f728662a5cc869cc9844b1b6eb652..470c1152938c4c0063812eb237c322f559cbe72c 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_tr.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f0c205df4fb30cb695c2eb37a6261cda7c9bca00042d6e089f55ac77cfdfe2ff
-size 3624999
+oid sha256:fc9a88c7eb0bc1e9ada4c1449cd99ed81a51fb4d08952c09e11ce63e1874dffd
+size 3624384
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l2_w2.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l2_w2.parquet
index 56e1d50bbedab4e7813a744d004e4c3f0830a1df..0f42225e29c0708c90bbfbf79553ee327e40788c 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l2_w2.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l2_w2.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:58856c675c2836ccb2e157c694997c1d116c5663e00cf1fed53c4c885afea8a3
-size 5808730
+oid sha256:e3f62a48d5a702321c854a08d3f19b94e1ccc027316acaaa4a1e5541ee885ee0
+size 3903011
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_tr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_tr.parquet
index 1a2ef3b5db356a5e57eebbf9c989f59e3d597f37..183c86178eb1ea782451e441dd2ca6ce6c2bf3c5 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_tr.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e0afc151e2f9ecda782204498527a3ac2e564c99ce385fc0ee77b13897dbcd2d
-size 3624916
+oid sha256:def63d08430dcd14044eeaca0dcd48524d6959fe01f104af8467ae871f3d325e
+size 3624302
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_onestop.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_onestop.parquet
index b16b7ebb890fa75d72356a2457e0c0c028e23199..b75bee96c53633e6a074f333b39d7a099915e15b 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_onestop.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3f156a3588a56e48f292970a5db0998c119392099e87c51f330bdacd1d1910fa
-size 64584179
+oid sha256:c24ec54505acf7efc1f885e6283572e9a7342c72c1b2eff106a380e4737d0088
+size 66298912
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_potec.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_potec.parquet
index 34141d5f13544ab6ddd3b0d53ce3d55728da42ee..81aedf9b079ebe3f694cedca4e79dde60a77921e 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_potec.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_potec.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5046abbbf409074539f9b5fc4806540ffbe41bc754d66437b74bcf09242d92ba
-size 3969149
+oid sha256:f0c0fe3080bde90f702c4d27954326f4435dc99ece5ddb62f58fdc6a82f6e723
+size 3968172
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_provo.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_provo.parquet
index d49e7c2bb776e3c88e86fba09e61d2a0ac0aa12d..417b9338b02b6e26a79d3382234a00c4e930ef9a 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_provo.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:325fd177203ee965f015d1508491f7d45e276a268c8e490190a89a1cb583f6bc
-size 5388611
+oid sha256:c856a8e8aa2da95169eb854d6257d2271b925369b7e6f5baa33a7cb5926eeaf6
+size 5478389
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_sbsat.parquet
index ca9817be72999b6aa7faedd3eef5bf8922d3e664..c65b24190454c1356632e7972fade245a759c92b 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_sbsat.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:174d3efb09f5b409cb55dcf21ad183559bffb604e6251bf47f354da43d626d57
-size 4997096
+oid sha256:91c1016a2f11bd41a27f1e008c486a2dc545d38bfc68622c1cfd7a1d7ff687ef
+size 4994084
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_story_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..84b55f6d571cea29676c8e0dc6d6c7207c27a04f
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7abab5bd53fe32ddd9a66a1992d6e8616d941dfd89f23e9df294edde54dc70f3
+size 4994106
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_celer.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_celer.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..0ba25d1998fc1ad4c9bccd1e54f1b3c16904d34e
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_celer.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42b4712423f63476175d129dddaa39c7a29e8c459613558ec784d6663fb56d96
+size 798196654
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_copco.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_copco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..c1fd1df2763123918a2caef4746e8a48d71c229e
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_copco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a6fd246d0a189592ca206cf9195efc6f4ce3dd2a0b0acf6cacdfe1ecf115246
+size 99760365
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_dundee.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_dundee.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..50708d86586e9b31e029c40348356f502621717c
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_dundee.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9bd88e751928100b4a48f3e86ec616b409858b5d0d9610c51f9ff54ff53adbbd
+size 121373606
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..fd97265558df8ba966e71a36496e0fa7899e9f34
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52815617bf1b00c8ee16a42f21e399f980f7e4cb16f5eefb99a2d735a541c9dc
+size 13280870
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_geco.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_geco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..e926fd19a6d0b865d566b17e27c0e3d08a69b233
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_geco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45c943ee183702ec576c6faaecfa5d41700655660067178f68dd1a959a7e8fe1
+size 146937229
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_de.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_de.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..ec7a6a46a72c2abc855a81939ae8b184151d123f
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_de.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:403a24b39690646e5ce03e363c2850cbc3309df6cdb5cd2db30b248df81ec7d0
+size 6362074
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ee.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ee.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..3f4132d8500b0816da67bd4a6b8e6cd57e3d6a5e
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ee.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4793e02bf80b6f276825aaa0ad44b53bab749510765c16ff52df956bc5881785
+size 9030821
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_fi.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_fi.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..0a07ff2e201c0002adb270588b89b185de763bf1
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_fi.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98b301573196bded24b21f3d9812c542891c73184317d4ad977ef9d13286f5e2
+size 7576118
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_gr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_gr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..a0da965342bee6c158ef026e865e95513ad0c323
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_gr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:184eaee6aabe0593e2dc246a8cbb72918b067fecf2dd383f7d2bfcd438760c27
+size 9009928
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_it.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_it.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..5b190d24e10b66edb7df2ec8527e4919eebf40e4
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_it.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2bc231359d6676fc38672a5e7d58657f451fb593b433f5fd64cacf6327f2c78
+size 6466473
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ko.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ko.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..f655d6e787557c3d364efa67ac47ecc6a68657c1
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ko.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9099203fd9c90130e1f11716dd89b4406e73f29fe89ff75c43c09d0379c2b8c
+size 8737433
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ba.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ba.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..20d44de640b25b3a02c447ec1252692ba0ecb57f
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ba.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6389b5653db87972d61dc6dbbe1f6921be466aad71d6f591d4f7cc488f17ac89
+size 8009823
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_bp.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_bp.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..4fd977af255e765b111aff24a9e4ef0f157e78d5
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_bp.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11e3d2b71bcf8363aee49211118c8175fa590b6f524e15b77c7890bd17f7a957
+size 6483266
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_s.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_s.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..2a5fdd8351c6c976d275822679347d432afcce31
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_s.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a515086d5754f33ebb59259a10ef5dd95ae391bf8ccd37b9796f0947d102ba2f
+size 6464559
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_t.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_t.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..d5a69e6a560df0b66bc22288452e31546201f259
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_t.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa66e91b69662a23934a9eb02f1e4249970c8f3ee9a037755530be98204b3beb
+size 6355362
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_po.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_po.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..df07c84fdbfb5ad7fde1c441693f229fb65087c4
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_po.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1f7e02a1eaefe84feaa00954b1e7ef88941756cc655f0c563ae3601b3c06db4
+size 6327709
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_zu.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_zu.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..c16b82e52a99667bd3158dd51b872cb7ab509a1c
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_zu.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:551975223ea723eb285cb7b71a674fef61e8cfc6da51050c262861115a9efd5d
+size 6327709
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ic.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ic.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..109115681f92fd62efd562cd93ad9899b3d0ba3c
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ic.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a512c28e0878d766ca8b49e43b95cc4b6821a9619f24005a0a0cac28510e7ed
+size 11098148
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_tr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_tr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..7cb7fee9e3b2d600255dfdbec7fda73038fffa10
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_tr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22144e2974cecacb31989954ad2bdc886f21bc629247f93e86c49eeb5c845918
+size 7566807
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..622018ad98dbd1c93294b562404a7dd830af2e6a
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dadbb2ff539e4922edaef04174af36a3e5c29f0b19908aebd6d5c310d8fe8661
+size 5010276
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_tr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_tr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..ef376f93556854312614b7affb43e1ccc612ee58
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_tr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2d23ec2fe6da9ce67e586416356cf82a060311a85a407e0794f925511537879
+size 7566725
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_onestop.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_onestop.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..5ac59b5f234bb6e757cac41c08f0ff077496a7c5
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_onestop.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ad2639f59d949c74e7507947f3710041711a1f0123c88d5503a735ed542febed
+size 87105351
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_potec.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_potec.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..69f3faf9523c202242e9ee82d6f2336387ca6410
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_potec.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f178cc7efd1e5e7625f12b261951ac5422515e1fcc16b704c8c52f0e6922bb3b
+size 6556196
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_provo.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_provo.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..52fe391eac609201fe382b36fa47519dacbbe2bd
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_provo.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cff170816a08be9244bc04f697aef769d7af4563ebbbba97f611c150f78ee976
+size 6921125
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..d3334c07d023adbcb3587d071b414e8d9463f163
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8618cbded2a0510afc9ee81b3faaf1e3c3707f4f9adf7253cf8abfe1a4e61cc8
+size 6462575
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_story_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..590d081a51f6a68b40e7442323d8a298dec7a2c9
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a48acbb0dfacfc5afb4cb8a97f3e19557ab4d19f3bc8e2159b20dc8e6442c1a
+size 6458641
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..697698515c0e30d1e830b03828d001ca9a397b01
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09e6b88cf1b05fc34ffe39d237d9232c5d4cbd1b88898529bf38075d36fb90d3
+size 7254610
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.parquet
index 6bad20c3b6cdf5899df1786d157ce25a51326fc5..ea9f8c92a2b42655a8b773ff898e7f08d26a6a1d 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4f2dd3e3a9ca43538c6fbd23f2eea5d592ecfcc7d3bb2dfaa70fda853c264016
-size 6409291
+oid sha256:4c6f3e124c2f66c2d0bee4c2a16d1b540e55e6506132df570a803ba7a3aa5523
+size 6408747
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_celer.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_celer.parquet
index 8751f22888745db447392a4ac07011d85003f5a2..fa8e32c72cfa52a2115a4a9d2021a74a040b0ff0 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_celer.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5633fc26aec0bc45a2aadce7d2bdd0de0061af0d417c6c1302debf6cadfc7355
-size 73357009
+oid sha256:d258093fe4d3d7517a58f9edd79bb9c51cb91c91353060c634a2ff3a2acfab5e
+size 73361153
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_dundee.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_dundee.parquet
index 316ddfdc1e9c917b1a96e6badf648e0f0c61fe7f..541211941555d49afb926ee6ea3130b4f5c4cf26 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_dundee.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a6ef731cd643aecbb9e8e255266861ae5bbb2a4c30ad759b9f33c49cb3064d07
-size 12158841
+oid sha256:4a130417111f9b544ef79439f35232aca63a53e646e03966b1beed9554d5b0f1
+size 12352630
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_geco.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_geco.parquet
index 66b39c43dcdff62c7e35cf9913e9a6c22ddbb961..7fe9cea668c96637de76fba8e8976d152bce47fd 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_geco.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f16bf6cb12df52b1b68e13495e51184ecbf082adcf07a86e4306a167761a5d1b
-size 13432451
+oid sha256:8216068a2b52558abf117e9c22855bbcc11491192dcb34c968400f76ac04bff4
+size 13431759
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_de.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_de.parquet
index 448ebcad816521aed1408f547e0ea63ad4fa1500..7316552d2af3b06adb148f349e1118ae07069446 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_de.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:345a6dfe1076bf853f529cb430d7ef55cb966c863a81f8a49fced12f6be48b8f
-size 741044
+oid sha256:6cad05a36e82787be3fefe11ede87974996bfd2fc678a8d446fca85ae1d74447
+size 741323
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ee.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ee.parquet
index cfbdfb8c3681beaa0ebab560a4b74cac2ccf97c8..a6b847ebbfd0588c9a1a941004f628fea1a2bf81 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ee.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ce92fee8c61cdd5b79a33bbbde44623b7a544614d5a1ea6b626a13215b341c8b
-size 561724
+oid sha256:afe38370992ae7abb35aa6e9717bdf141f5335d041193e661d3e7a9e706b7c51
+size 561105
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_fi.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_fi.parquet
index 8cf4a705f88eacd21aa75277193978802bdc14ff..800ea982f97877eccbd41d78a4acf8821dd92403 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_fi.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:228bac799cbd17654bd377990131094754232de81142aff719181da0cb8f2b51
-size 545731
+oid sha256:e26e6433f2205823689faf10235412b76c47179f6ed879cf6fa86ee63e513a35
+size 544597
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_gr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_gr.parquet
index de4650ca25364d5fa31e3bfd00ec0fa086b17793..b4b94cd44c47789976dd066bff343e153b1530a1 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_gr.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f874a2df6bc2ffe641a5d60d1158ef1ff1bf6466f18b2fff1dc67d009d90cf73
-size 765653
+oid sha256:c332cac462e64642d92a007c315fbb58cdcfbfe0e04f2a5206857002bf476a26
+size 765293
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_it.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_it.parquet
index 553a58852846419e4708386fd347400b72ddadce..324808802005c43071165db8b069b94bdfc99d0d 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_it.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:368cbdb309f57a81d49ff054f63f11f8819fd0d5f3aef50395ab69258030a8ba
-size 784678
+oid sha256:f9deb803e126d55579a5494901dce8535f1b331fb413edb89d1520c066ee25fc
+size 784146
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ko.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ko.parquet
index 54f9cdbb4c528165764e59cbf22277fd7f5e4a33..022894fbae9f767487712ea6f24f677f8b28d048 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ko.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ko.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e2a24a741cfc397a962f2909450624106e47c3ec94c1e3e01df2fd41701568be
-size 658340
+oid sha256:e474dbdf191dc0a883537b454505517164c03641a6d73ef7d421e78f9320f67a
+size 658434
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_tr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_tr.parquet
index 38c6d2543a9c3138bd0601c96840cf9bdf3ed848..100ed774ea9545a13384fef8bf821536033a94b7 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_tr.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a7f6b4fb21e4c7f21212488a71302b87b36a6cb607c961927bfc67ea1fea1cf9
-size 621693
+oid sha256:72b8e232f569e88ade88179bb7374d79e87c173e318c979650d914933d37e5e9
+size 621359
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_potec.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_potec.parquet
index 977cccfbcdec42b7950f412033623f0720932e4f..f0daa301def3cdd54a183366bf8423398f5ef0f8 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_potec.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_potec.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ac9af13334b789be341ef6685af9f95daf944ce223f4f06927ae42b9d284abdd
-size 693271
+oid sha256:93d6baa063c0f00b8c9a53c9a9212ff2f8a33402a6d940dd64de5397ad1dbba4
+size 692987
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_provo.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_provo.parquet
index e90a8f00d86e7e7660188584f800aa3485274b93..c4fde2d86df86471a576d228b9a8c0074fd51238 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_provo.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3309afd50de1d3c6fe923610ad4e77ce90b9ea63b898d1afecd8311683518fa7
-size 993751
+oid sha256:07eff418ba36c8f3bd5672a84188981e3cda637acd6bae129077ecf1a491279b
+size 1013120
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_sbsat.parquet
index c99fe58b2b563f8fcf97ffa854f9d0f145d53d87..0ba172ea4e08804da0c76f76265f2d8f4eb48875 100644
--- a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_sbsat.parquet
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d4b439c66adefadf3e702c6d2e4aa9fca3ec00b12304d7ca9ed743d31f9b3fd8
-size 914821
+oid sha256:050ed17407411baa41010bf5c36e46bff65f3a928191ccb6a940ac0ae06a4a9c
+size 913804
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_story_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..03bdee37769f5ff1a45df847b8ebd0233afdea6c
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14991dd852143b243cf96e26e776eac6f5f52234b22c1c22b6a7367431321702
+size 913761
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_celer.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_celer.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..e4635adb1d3073a9d4430f89971906b9f370b7d3
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_celer.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03bdce8b6b2638276ec77a1d04a5ec6b020a6f18c04afc7cfa45a0da5860c469
+size 103531020
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_dundee.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_dundee.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..2a592b121913d11cf43f535e1a2491fc2469dbda
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_dundee.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fee0e339dd624666ca16992796dc465e51ba27c69d54e19df92ab18fbe61fe2
+size 16548909
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_geco.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_geco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..49c24fc236e51e7b046e1ec695ccbe7efbaa6e3b
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_geco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3151b1558456c635e63a7f2cf941c0aaec2e1bf50531741eaf7bcdfa2459411
+size 19637907
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_de.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_de.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..017133424c7d5d74fd7ff40304b7f498c5acc223
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_de.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72c8e0f1ea79cd822efd8b4b0b0ddcc1cf7c1b6ed64db657a760e7b8caa9e56b
+size 1212446
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ee.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ee.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..8d1c03f155d628c3db994466501cb325ad458429
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ee.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dad899eb3708ecd8316dab8ac6719a7020a7da9f28c7cd38c0ddfed38279d4b4
+size 1700740
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_fi.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_fi.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..a99f30252703300cd1d0d3951fd087b0c83e3a4f
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_fi.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a820d5500304b685c8cc7258da0df764cd272036a3ad195cc13cce8848161e2a
+size 1466680
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_gr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_gr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..aa4f23c1e1708f67cdad304e5055f9b78392c8e1
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_gr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a28709005eb780ba4dc6c68347540b90bf6a586ab8467e988c584e80688e93ea
+size 1730293
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_it.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_it.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..a93895ce9b49b8505c474184d3c8abd8f7bd99ba
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_it.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b0b3f573d953cb2c3f796dbcf7202c70b7e959d1184453442362f9b32ac9fb1
+size 1228091
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ko.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ko.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..05db5c913370962540b570efbd5cabe3e3fadcf5
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ko.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3dd7b40115de2505433c46c5113573717371bd0a69568c1609eb01c10528f0d9
+size 1676830
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_tr.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_tr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..1a8d9f167e457867892edf6377c900932b196048
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_tr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49f5004b3afc9e078dfcff893c4548443dbdb3555a8a049d57eb91df97906214
+size 1459115
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_potec.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_potec.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..dfb440dcce07b4c9bd4cc2ad3cdecd0ec5d65fa5
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_potec.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:103e93473709b907f69e4f0c5a8b7e30a72db7bb8adef05fe6292812b293ecd4
+size 1261072
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_provo.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_provo.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..79111e73c8377cf11156c0c89746fce84c527189
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_provo.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cdf397dc741bd99d70c3593209e0c93db5d7e766429d41601d1dcfe29559a60
+size 1325152
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..c133c23f284a1a5917243d585faea50543323471
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ef53a943f87a012d813d5e98d5f56df6c5521f3e5cb94792f996a4d695b89e7
+size 1230690
diff --git a/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_story_sbsat.parquet b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..0f94af38dc1561a549f672e9dd99c88fb7a4e37b
--- /dev/null
+++ b/cont_entropy_samples/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b8ba653f994440a4ccab8861fee0fabac59ba472e2652a46a28048feb730ce1
+size 1228044
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_celer.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_celer.parquet
index c5c5ab05cca8c4fde06008b2203c0f9bfc780897..03d2437529fdabfc28774cf2ad5417bb670f8dfa 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_celer.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c5dcee2d143c7720c2797b481f9d8974bf17f32fe9555372aca75eb5e49432d3
-size 576085907
+oid sha256:b9adf290234e7fb3ad1ca8f516a05cb068c277f7e65e8ef9c128e5147fbe3963
+size 574913490
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_dundee.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_dundee.parquet
index 31871ba28ca5dd5a9a461c0d5d94d9c1f4aed0f0..a3a8346192b2601674c6d0ebfc62ca861e4f6553 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_dundee.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:67cfa9b235e589a3c71e1fa1cac8fcbab5885f572779203af1cdb21c62197cf7
-size 91048851
+oid sha256:fabee5756541951346a26813df609672b6e0d1fca4355947aaf850a2b467cb74
+size 92724385
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_geco.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_geco.parquet
index 2a1319d033d1bcb42b9675347fd539581475421b..3746ab4eb986a8a6699ce776ce4c0ec719ef0dd4 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_geco.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:934ac3a1b5df7ffb9ee5feb348a1e5e2f8c01a81e1a0de680e31ec233f409566
-size 101750376
+oid sha256:df6b2cf6ee2b2203ca1462fd04c8f5b31f9ed02b04fae11e2f577af4c86a28d7
+size 101752752
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_onestop.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_onestop.parquet
index 03c33ea81a151e1eb6c3f914f5cecf7f67401699..7becdcbe9ec122674af4ba18cd7c541b53331fbd 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_onestop.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54eff7b5ee92eb8a5b7ffe287b1677f15191e08afe16525fc356c698a9d75d7c
-size 65393566
+oid sha256:232ca503c0fe2d8d8d14ac76957a4b3a7796425a018e2018ba33f16f6db341c5
+size 67127769
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_provo.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_provo.parquet
index 306b0832f88261a2613f84b3af195d4e72c96508..eb6a88829fa7567254c89c3171aab8ca2de9e1d8 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_provo.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:983753ca905b65e8ad82259f9fbfb376d8efd7979f1b22341cc0732c6f2b6040
-size 5440903
+oid sha256:f6a3c4d3f39edfc0e1930f025d911cda79c4668304ba534810fb6f59d16ac713
+size 5545890
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_sbsat.parquet
index f9c3fd121550a8a3a4494eccb8aa1c4db9743924..9dd1a13036de3dfc06f9af057856ec0bf019283c 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_sbsat.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3bfc6d787ee0d16e72e77010499ef31b36d3a1ce184257e2bfe504bc31b3a257
-size 5046807
+oid sha256:fb76592b21d2760a106834d7762df0bf84ca9e7d73676e2ae11ba5d948f59197
+size 5044739
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_story_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..ee5559218149d25ff64ea7f94809b00d712c2023
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:980eacfd05f08effed8ea0d87aa183bbb289f599744bb872aac3f2f577d4949e
+size 5052800
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_celer.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_celer.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..fa79d4ec6a877f6e14dd656ac6b21924dadbfc7a
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_celer.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ea797ac1995e13e77dca0a5fb19f7b5db03de72d7bafc76798cb70a9eea384e
+size 746127366
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_dundee.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_dundee.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..001ac316a035bd38bbf43aff14c556a6ae71081d
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_dundee.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:747ba83d0dfe165159cae49329a9e9aacf04dbd47165c81f27baa7630ecb94f8
+size 114192765
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_geco.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_geco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..f2c5fbd192c8e15dc898248b07f65269cb7e832d
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_geco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1616c0355ee2369dc32c565a673081cbf5ed42de6a5f9803dba24ce5bf46443
+size 139323067
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_onestop.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_onestop.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..2b8593d99e73dc4d49cce80b7f2236fef13c5a90
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_onestop.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:258f486d3853c11edf6119eaead8852547245bde936b691dcb7dff5bcbbe0407
+size 86305962
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_provo.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_provo.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..2c447609a1ce90460c6975fcdafcf88b457edaac
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_provo.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb5d1f704edc435f51b6c0553a11c832b1b68afdc4bb28587f73b0f7a2da5507
+size 6517416
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..eeec13da853c6054a9f15359546f87b97f54149e
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9161a1bc37854ebdc967d2a681c55767e989e05e63eef2b1e0ba2c2d55fdc852
+size 6197644
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_story_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..ee0812826f97fb5d680ac4313ee0845ae237db4c
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L10_s0_tokens_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fbccac5769968d22925de2e368ceee31c1ea103d5e8109e97e5b5ead1156c2b4
+size 6197290
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_celer.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_celer.parquet
index 8a892abc8ec31e84697ce27b4d27abda57e64cdc..c7d7016e85eb5a42a46a9667eba4f21b66aaf897 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_celer.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:51087f8509e10f32aab68353567a5d511a98f4646426dc69756ccfa56d77f2a6
-size 573049716
+oid sha256:802d8e6c1ff59d9297a1547b8ec519a3c31d3c75abb63ddf0052fba5f40fdacc
+size 571746888
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_dundee.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_dundee.parquet
index 2adfd25614a8cffbc10ef0341ec94bf67825260f..c03e01f7b43081c0c391da04bf83cdbbb538c865 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_dundee.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7601fec7d038ff178e9c1c8007af7187d73009255b467ed6a0255d652f380483
-size 90266196
+oid sha256:38667d92155fc28b07ebeded982d68d6414b6ddd82a8c00d8d9a35278e0033de
+size 91764968
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_federmeier2007_n400.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_federmeier2007_n400.parquet
index 1638af82679f084a593d31db06f2b881101f85c9..3ab65c0c3cfdf8dfeec3cf2e036ef0f418a598e6 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_federmeier2007_n400.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_federmeier2007_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9ecb80fd7d683464b55d4dc2b98650e1a54a2bb028c0c088c8f7512219f4df28
-size 11548401
+oid sha256:3978592a64cf130e6cbdd00b7eeabc52cedbc4c1b1d876ffc86545018df43a81
+size 11548646
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_geco.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_geco.parquet
index f4b31923f27f97eb1107f24f62e7976fb7bccbed..03739173ce031534c59f97704c00f927b986dc0a 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_geco.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dc9d444d4921296e7ea95eb459350de9e836423d9f171e6c998e85982baeab98
-size 100573660
+oid sha256:4674aa39ec6a76a9b9feb6f99fb0c910e4caed519c64c54b7be1e68be485195a
+size 100570531
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_meco_l2_w2.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_meco_l2_w2.parquet
index 0dfde5f64bd7d2d2cba472e3938a4f12fb567c6e..a4fa26d01b355362edaa4638d49f4011e3677a3e 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_meco_l2_w2.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_meco_l2_w2.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6a36cb6fc09f689c818361c4861917d108b0fe2fed4f3a822fb8db1e62fb40cc
-size 5796784
+oid sha256:2c2329912e5797e79c2402564eb28f9113db44843bd7d073196c5af6a8fab353
+size 3695390
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_natural_stories_maze.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_natural_stories_maze.parquet
index fc2711bff1e174f8066eab5c1d18beb72cb4c45f..f695d5f1a83c359addba61af630c89dde7eaa69a 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_natural_stories_maze.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_natural_stories_maze.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ab06996840e2c02a0cb8fda4f7920a170b6d2d5b2fe1b4e2bd1156011f0ab51a
-size 18812170
+oid sha256:29dde32efa063808d4b5c49ee5552d1529fd82485eb336e667055e4622344b76
+size 18815911
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_onestop.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_onestop.parquet
index a58fe64b5575ddb67f58c74d0d09826c21c44290..c1d2e3d901895dbc0c6c82c211531f7aa2bedbd0 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_onestop.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4b83b6facc1531219309935d7036f93243a459fb247828da6f3da7a0a64d8164
-size 64593671
+oid sha256:27e0f814f2fd66d7a36f7ca33222c44d8f1b9b94b300c12ad52d3c6d94ab40cc
+size 66302772
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_provo.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_provo.parquet
index 40befeb458d0d07b6674494d4c1ecf2abf90c3d2..222c572eebf67f6c502e300a94742f57d2a560dc 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_provo.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6475a7cefff6e43351a86df1ced55d3a5ae2e6fbf15ca519d78b7d771ca47758
-size 5392115
+oid sha256:7c96b545acaad49fafaf7b4a5809c92cc3afd800e085fc88a1a0c61eb7e854e7
+size 5484960
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_sbsat.parquet
index 93f6fb849ef8d842e7c3230959b0cf0ae980b964..bdb484b79bba23014dcaa4eee7d0c53c295ba811 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_sbsat.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1651020c28d923f26955267d2aee4dccd8aca1bef931ef7ac9e12679e5c49830
-size 5000258
+oid sha256:54a8f9b8029920a41d21291b14d5c1f2f523e09803485a1335f46d446402091c
+size 4997937
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_story_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..b6a6e7607791eda0d7b24ec4b146a387eee1c8d3
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e32dbfd91d2c3d4c52a469f1c48d28cfd79b42185b5f973f724074bfe856b21a
+size 4994382
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_szewczyk_federmeier2022_n400.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_szewczyk_federmeier2022_n400.parquet
index cbec96fde794d1c0d7d9d210bf400fdaa8e66054..48011587de2361748afb09884b2d628667379917 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_szewczyk_federmeier2022_n400.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_szewczyk_federmeier2022_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7865cd6473e94135a37eab1abcebaaa43a54057357e02cc4d3119e15643deb83
-size 12155637
+oid sha256:b0bdf6ccb207c1e07e70cf9dec452d7a83c847044b5271fcb32651cae6754eb7
+size 12155568
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_szewczyk_n400.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_szewczyk_n400.parquet
index ea8e786ad27e3c6a0506dc6ffbde0d2159c5e1b3..51a86263469543244a1f1e1139cc14bd18abb23b 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_szewczyk_n400.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_szewczyk_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0ed309764ffdceb4ab308caac3a9737793710644fefecfcad8236e05c6fc4152
-size 27000555
+oid sha256:bc45428fc43eade0706106d2a415e6f6b52f4e4882c7cf516d93deac087feb7f
+size 27013476
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_celer.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_celer.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..6d17759a7b51eb09a27822c877de9361d9abf67b
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_celer.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a796355e574363227b02e5852b49078f95abc2d153f5eb957fc0537dc3fab2a
+size 741298283
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_dundee.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_dundee.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..49b81956f1e1b0a508b86913bbeb36e87ce76cf5
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_dundee.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b94f9bd709c29f27fe01ce824752e4c6c642062715664ac633b4666f64fdb4a
+size 112997175
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..77eb67ebfb4a17bccb81ec905eb98752e575f7bf
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3c0ebeaf184c1f518edeba41030781f9f835dea738d9842f6589fc1c244a4de
+size 12219110
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_geco.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_geco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..48971c098f2dbaaf108c162ad26a101ab3b10308
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_geco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1dc72d197f162cdc097d7107894265ed4602d371feb22a851ca7acfd2ad36bbc
+size 137643137
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..03ab909e56e64c4460b33e3119bb395a261bed8f
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2293941d85ca3c7c3987f6d2395480017763cd569d360128a97d7c6c55796c51
+size 4520154
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_natural_stories_maze.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_natural_stories_maze.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..e57a68c1b6370688d29fa3a44fae5f4c8711d2fb
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_natural_stories_maze.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fca63fdf84844b8644b7e2515ae4e30586a659831e808cb63655a8c21629180
+size 22700852
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_onestop.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_onestop.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..796c702c18b49452042e37599635db9c769bf45f
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_onestop.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:292b125544ed5bd67fb39fe22110a67ffb28e35a7761b26ce18197f14b619778
+size 85283372
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_provo.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_provo.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..d226663e9933222dec7bd8a8165921f1c36dd52e
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_provo.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3174cc821fd3773ed2b3ffc0a78fe48015c30412f56f29c15cef888b906e4987
+size 6442286
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..0247dbd24122ca1ea53a74b20f63a7745541bf57
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df64586298b1ae0b4377c06befc88dda875cc7b288bc327310da75117660df73
+size 6139505
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_story_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..8d430c78afa5043f944e06d4a671ea9860769e5e
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d564e8d84c88c72f181b480f658e9da52b4c0b130ecf5a808739d33ddb39c770
+size 6127189
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_federmeier2022_n400.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_federmeier2022_n400.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..14e179fdf6de474b9150fa57a5f33e8bf1262a04
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_federmeier2022_n400.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9638d70778646dea363e0f53abcf99558511526dc7dc792a4c5c52c3e29488d2
+size 12834624
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_n400.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_n400.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..f9e605b905dfc5be23de0ad3c57422289b99f69f
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_n400.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fda265e7afc8729fd00476d135074fbd83fd839d1d1e59a0980947191fcb0235
+size 28664920
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..095bd0399b2674ff64fc75de40633ee07c9e0636
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2a067a38ef2a59dcf1aebfcad0c86a2d792c156fcf23cb595663478ac325526
+size 6732751
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_zuco1.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_zuco1.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..5c830b7e298db40562e1e767eb0ee057515718e2
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_tokens_zuco1.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ad05f22cbcfb063d433c0a52b4ce3fd9c3d334a64e800f0db6604b692685040
+size 16325612
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.parquet
index 9c78db890bfe5812b5f338db0c8715c87c364f86..3f4190a51e298559e3f9c849e2b041444813acba 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:38272e6897dd7bd318480a47377c96037a7aafd3752280a7bf86e745933e1fbe
-size 6412966
+oid sha256:42a05e3aec0b90b5190586c9852653584e302885ec38a4f9f8df806ebf13b09e
+size 6412680
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_zuco1.parquet b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_zuco1.parquet
index e379ac14a9209b1a551691fbfbcc9f71c93e106c..15eadaf6dd12831e03c9855813fca22babe1fc86 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_zuco1.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N512_L5_s0_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f5912bda205ca42737a262db42965fa640d82a519b38adeda15cb7deb3422863
-size 12347809
+oid sha256:d95ae3735c0b243cb5c9e9e6e1bb9d5c2ffd3595842bf427c2f2797c04226869
+size 12349716
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_celer.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_celer.parquet
index b94e34c5bea10f2dfb4aebb4946a4d8497cbaa32..a89b88efdefa5871926fbd9ef1c874a3bfbb8fa4 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_celer.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f70ea11d28cde960c9e629a181eb999801b010615c52266228a8fab8443f3901
-size 73374787
+oid sha256:6aacf3fe3b0d65a9191b022bf0ef1e8c2946d53d0d339cf8d86c0fb19c7eb915
+size 73396977
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_dundee.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_dundee.parquet
index cc4e46e13ea534c30992d74d65ae30b5e4f97334..b86a43f950e39c7d3ed3e9a16d0d703498586fcd 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_dundee.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c1f0e8179821e9ebcfe49feb763c1b702e18c749fc08c01847f9c3dac3090b2c
-size 12085381
+oid sha256:51d3fa301289e2832946cad7091dd4c2bc33d68efb618556854580a1640d4824
+size 12298931
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_geco.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_geco.parquet
index db89accec9a79a4c91803440d8249c0b21ba2ab2..a80cb48d0c814d92cf8e00fcd8394157ea51a3a3 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_geco.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:673fce6c3e057fb04e85bba78c4c888536f1e9c7ee2fa232e82c1a0f87d3202a
-size 13397035
+oid sha256:c9df33a6d0f2e652b9aa86a0119c1c50d440dc0b76edbb4cf870a1bbb7ef68d2
+size 13397195
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_provo.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_provo.parquet
index cd5a28f8cf9af3ee6d4bb72540afa797cb1c7fbe..99898591bfcd174bd64fa1b3c633dbaf054eb2b1 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_provo.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e544fdaa9d5f5c841bc346d06acbd27d6acd96df14d994f1c2d822b41055b69f
-size 991594
+oid sha256:6f418312daebc31d27d89f1e360c1754e5f810ac4ddca66e394b811b22f6f052
+size 1012463
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_sbsat.parquet
index fddb61f5f3764dc3afeabf0fa985364361035439..a277c68d5daecb28e8bb22230ed86a1d4c6b5b07 100644
--- a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_sbsat.parquet
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cab9bf7cc12ce8a73c03c59591877354293d082591f06d4bb8a9db905f254e2c
-size 911304
+oid sha256:fd97a8d195f6f4cd46317cbeffbdf4fb9a3c22df7a538de8640a9013e51f1286
+size 911274
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_story_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..3a333b9f64ed6c101140f5600f76b470c5d0efef
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19a5517ea1e1dec3517abc5e94f7cba2d46361625548c4265252bce4279d1d90
+size 912803
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_celer.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_celer.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..89b23f29fa6411d5cc1bc92226c12273d055147b
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_celer.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b021861708630d79f5bef0b015c39fbfd97098134b75b936ed79210e62dcb78
+size 96135396
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_dundee.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_dundee.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..79ac08cce10688450c7af120fa064486c982c6f9
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_dundee.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ba8bfd9a56240905d783f2c20c76925fbc43331dfc164b8c47df1c404abc1cb
+size 15392364
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_geco.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_geco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..2b13f538ba9443e96ea08b6fbd1680855a28237f
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_geco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4693d6dd8af8c1337499f36bafd8807ed30e8f51df33b697ae53126ab9947ff
+size 18390829
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_provo.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_provo.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..b3e48e6903ebddc76c1e44b57057ad580c2fa9b7
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_provo.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56c2073d7d87d6e463fdbf749782875c97be3930544f6975dbf5cb442841a6c9
+size 1224060
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..4c046908d14dac14e93aeebe5aead6cf6f3d2e40
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c4592283dde44f9569ac68009bffa4e1963946713ce2ab41efa75a1bdc43bd4
+size 1158433
diff --git a/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_story_sbsat.parquet b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..2c675b17f69326d557f16b6b4b61e88945ef371c
--- /dev/null
+++ b/cont_entropy_samples/gpt2_cont_entropy_N64_L10_s0_tokens_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1130a3a8231dc0feb9bf3da8c848e3c5f6e01dc4263d694ba73746e7984ed00b
+size 1157263
diff --git a/copco.parquet b/copco.parquet
index 8ee10add3d41f4a57a797a5af5cb667db3a81df9..6bf5ac0e72c9107232c9243d791cc0bd7785af68 100644
--- a/copco.parquet
+++ b/copco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c948df4b5c63d6c85934442f6b86cb79ffdf2d81eb6b1dca0408365f85831c46
-size 4343011
+oid sha256:edc4a07c405d015af8d8c9da7ca0ec1edab9837bf19450a06876d64e73452320
+size 5976780
diff --git a/coverage.parquet b/coverage.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..a15d5a73b808c70826b960adb572d100410fa5b4
--- /dev/null
+++ b/coverage.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9be70c85eba239bbd47947709627a6ee84e70859790bb5b33bcc21ef900cddfa
+size 16776
diff --git a/dundee.parquet b/dundee.parquet
index 2f51b5a1b52465012dc7dab72ebdd27b84a1108b..ea80ca1fd7b723b5fd880178103c48f4e8d2a181 100644
--- a/dundee.parquet
+++ b/dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:875d59f3678710b4ba21810c484e43dfd745a79f7d913ca7c92fea8829f09a44
-size 42656826
+oid sha256:d3b955bdc56e59053f025ec3a3bb3c81d3acbd608bd68f080d787c2bc073d280
+size 41711159
diff --git a/federmeier2007_n400.parquet b/federmeier2007_n400.parquet
index 958329c0f2ead1550ab64a970403b8afa1a5c349..8df9d7fdcaa68760506c8f5bd4f9da3e04aae400 100644
--- a/federmeier2007_n400.parquet
+++ b/federmeier2007_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6d675c136c14067e0819b8af1cabbc89a86fa2cfc3a82fe349e528879c2f00c2
-size 2295116
+oid sha256:a952fe2079d20cc7bb8f30e7be5446ec232f7d331a53ab13183692539653a310
+size 2539483
diff --git a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_de.parquet b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_de.parquet
index bfe446ae42649da59cfbb8920df7e0379b830b7d..670b9b14ef5d37d18cf732d436ad941766cd3cb4 100644
--- a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_de.parquet
+++ b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8be0d3fb0f2871ef2e7ce6e20619bab33eaf1631336f4af976cdfb1f7ad62fb2
-size 30374866
+oid sha256:20e6e8e16d181885b35c13e210a46651fd23a3b5d85d0071ea5c3f6daef490ec
+size 30366220
diff --git a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_ee.parquet b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_ee.parquet
index f7ef26332d919a9a7992ae41d3059837b9fbcedd..e8209234df65aa00f82ad2394f72afbeaf055999 100644
--- a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_ee.parquet
+++ b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e0ee5c8f7c2e22cc02e7ba260f888f7f0ca780ca3217495897fc052ff191cd48
-size 29800157
+oid sha256:6c568a657e58188497c376a6512b1e13c438cef3fb4131d44ceb94910e1be76e
+size 29781371
diff --git a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_fi.parquet b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_fi.parquet
index 786f231af6d1f88020659c1de471f3618616ce4e..327bc4aaef24ea77d34013eee78f279c06abbe45 100644
--- a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_fi.parquet
+++ b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:83acef8597168b4a7aa4552e4161ba475bae3876d9cce483a975e9f1afe1a75d
-size 26552139
+oid sha256:b43935a1edaf8d4f73eb9f0c1dbaa0a43e69655170fe1d2130a156a7a63020d1
+size 26503333
diff --git a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_it.parquet b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_it.parquet
index ca9e3ffd07d28a6a3b3101bf62dae6fc21b7b731..73081fc66c9595f42333d51fd6c5c4c7054fe3b6 100644
--- a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_it.parquet
+++ b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8240f8a1f51483ce049daef9af7f924c0848462bce0d517fba524281beefded8
-size 30161749
+oid sha256:553610e836b12cc1368c67922c671566f70fae6325f7397553877e622accca94
+size 30168724
diff --git a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_tr.parquet b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_tr.parquet
index 00822dc5cfb7dd71047e49d9127a75d1b8dee64b..e44321400380b09dbce2edac3badccda3cb2e6be 100644
--- a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_tr.parquet
+++ b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e57d50ca658bb02a155c0c9e46ff54a2af5de0421addba325958883f424cdca4
-size 28419372
+oid sha256:36bac0925ca1496581306699821a8508a202b41bb533ed78b8e9601f4fc7eeb2
+size 28394507
diff --git a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_potec.parquet b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_potec.parquet
index b6e28b6183f63d7d23b2538b57168c842092998a..f243bc78ad6d956027188c2bb3ad3053ba4c55bd 100644
--- a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_potec.parquet
+++ b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_potec.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:89e45c3dc54a2ec4daf3e751b62f85768beda17f6d4911a4eb8d0e5acff004be
-size 29679721
+oid sha256:6913f41f38d1bcc02715fec6e9dd02b6ef3bb47b63833010b60c2aac15ab0f02
+size 29627971
diff --git a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.parquet b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.parquet
index 7f4cd36aa67eae6e7a8f67c9ff6af4a92c0c629f..9f5ed456cdc71a805f322f44c4ebb9c4b978bd8c 100644
--- a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.parquet
+++ b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:947b850b8911a3ef0b6aabb1bcd9d62188f8234273795da5bf35cf445a4b0ab5
-size 40594109
+oid sha256:f0af45cd4c61561ce9832bf9f7e23d2c866fcc7c654d72e9e31d56b71fb61c54
+size 40928902
diff --git a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.parquet b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.parquet
index 12de56a3dc29181771ae0f4f17d54c61af8038a7..bcba20b65e2120b8b6d08e2877bf93bb76b75771 100644
--- a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.parquet
+++ b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d72a1c50a8da1bf8513d5a91acd5df8a615da4edbdac446fb9c3fa945524ee2e
-size 36198638
+oid sha256:2299c5a6d1199abd29c6042bf8a4deea56ff5fbd7c264153661eda885653b46b
+size 36192958
diff --git a/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.parquet b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..14f2d537665ac688b0a19cc272ad7084d3f1e0ad
--- /dev/null
+++ b/fwd_word_lookahead_entropy_samples/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8108516e86781632fd1ccc97473e2edf644b03423fd93cba7d08ecacef44f540
+size 35058601
diff --git a/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_celer.parquet b/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_celer.parquet
index caed69dba82f87f1e43a506b5bcdacbd4f271680..b8b9ad93473d3cdf4e00dc7091ce08de0de29740 100644
--- a/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_celer.parquet
+++ b/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:941242bf571473c3e52ac3dd855c7a2f3bae38d2fb68b4e5306e52faf93d6cb2
-size 4673507249
+oid sha256:1789e2cbae9ea38e63510f6b3d281edb58a82feb0aa756b089825e381ea77132
+size 4645316838
diff --git a/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.parquet b/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.parquet
index c7df6cd868445e3c8cde535a6235c7d500e00efc..1cafbf25fb49465e6df9fa211b79daf0fdc8a6c1 100644
--- a/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.parquet
+++ b/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:52c2fb6f0979bccbac7723926717efc0a09903df0baeaeb1438e31f9b0ddac49
-size 38349901
+oid sha256:e18622a0119ac35713a0403971de541c444b97ec111e836c002268addcf32a8a
+size 38760167
diff --git a/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.parquet b/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.parquet
index e482df789def89c6bca7834dd00d13468a2f00b3..413c5eeb555e53f627783960308760f371a98349 100644
--- a/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.parquet
+++ b/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:91c44d7ead0bf12c08be426ae5caec876561707a586a38e8e147ed254fe5bcfc
-size 35454239
+oid sha256:c6df3e78d7b9f627d20fcb948065c179f0a24b73c6c9aacd04c9e00af3ebfb46
+size 35354433
diff --git a/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.parquet b/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..9a8e5af74a3bdefc139460cb987313fe54c1c732
--- /dev/null
+++ b/fwd_word_lookahead_entropy_samples/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b68bdc2e0232ff385d654bb4e387d1383c0ae20ef5c17a9c0c551e916c06723
+size 34509942
diff --git a/geco.parquet b/geco.parquet
index 1f43d9d2027173e7d64c73341485d58853976baf..4961b25126f86558fdd6393b860c70ef43268579 100644
--- a/geco.parquet
+++ b/geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:da795cce647597c41b43146ec63f8ab3a59ed79098e668b0717dcb90fd1ff973
-size 51670228
+oid sha256:0e2e81992ee05425d51962b6141885b8cda1479416da5de8b186481e34867ab5
+size 47410281
diff --git a/history.parquet b/history.parquet
index fdcb8966fda1956220968674c7875bcffb35d08c..645b6cb1d109d615765e080833bd0bad8bd49440 100644
--- a/history.parquet
+++ b/history.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:764eecadca37e69772b4d38cb4243b7dc5ea0d5ae47b977128314056af81ed68
-size 17324
+oid sha256:2d5361166e5bfbce7fc8cc7061fc47e29cc6789dcf25325a6b20e9698cd9d7f5
+size 29293
diff --git a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_celer.parquet b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_celer.parquet
index cc1e691bf380231f52d18fcca8c80ec4d6e29683..96444b782f426e08e2ac05378de27662ff16fe8c 100644
--- a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_celer.parquet
+++ b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:489c0583f85c35fc98e81ebf3e15026db09d335a99f7a5e3c48ef8cb49819621
-size 37336726
+oid sha256:380530447335f25b74fe52dffc4fb92bb739a22ca5d9155050568f4a523becc6
+size 37443160
diff --git a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_dundee.parquet b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_dundee.parquet
index 51e14507b8c64e1346448c11e3b3940cfec2d4fd..6d1215c8c880bd49ef174e6adf8245701ea18d10 100644
--- a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_dundee.parquet
+++ b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5fdc05f594c09d6c0ef1105d1ed3d5ddd9b6311d5d918393d256194469bee608
-size 6405408
+oid sha256:0574d155a54896f4f63e8ca5149b9d34f6317b57413ed8f5131fb9a33c2b780a
+size 6511260
diff --git a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_geco.parquet b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_geco.parquet
index c2211983d81a485e94f11d43b004d20e4fd3953f..18fc463398c1b7bcee967d945ef0eef0b9ebe2bc 100644
--- a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_geco.parquet
+++ b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6469774167dfeecab8b3b13acb6466570861cf7e6aca130cfbfe7ffd7e336eb1
-size 7087243
+oid sha256:5a3b4a67aa17e21aeeadc0cebecb44361928444a45498a9a17852fed36ffc067
+size 7093307
diff --git a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_onestop.parquet b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_onestop.parquet
index 59630c3b591911a5bce8ee63b14fd5ffdf85dbfd..667ced590ff3b1291ad766f812b619a19f17d71a 100644
--- a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_onestop.parquet
+++ b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ff57c91b1eeb7bae9d7aecfd04cd4037eef833a00a1265174553353850ada72a
-size 4553717
+oid sha256:851e05dde5b3dceda4a64d14be4564df711d7a72175ff063fdb6dba6ce5daebd
+size 4567748
diff --git a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_provo.parquet b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_provo.parquet
index 29c34d32aeb667137613614f6fc65c337571e609..6d97a12596971464641dc4534f46fdca5970749b 100644
--- a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_provo.parquet
+++ b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:354d60ee4854a1caf4f581d9c6891d98dc554270f29dac88d86d3551b3499603
-size 467247
+oid sha256:b5470805a6b198121dacbb1b256eceb3fba61b1a4a9bc188573cd130c9551cd6
+size 477217
diff --git a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_sbsat.parquet b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_sbsat.parquet
index e071bb2a0b452ae2cbdf118ddb04de95a3d0b4c8..0373a5bf340d50899975950c87e94a2b47d202a5 100644
--- a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_sbsat.parquet
+++ b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:53c7178862e4b32bf5e01ceec7981f6bcc5572040fbf39751ff4368914af8b86
-size 422948
+oid sha256:7b714d567a19e4a1213e6a6ff66ba81200c6f7bb8a81f6f894668e7e937f002f
+size 424686
diff --git a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_story_sbsat.parquet b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..d31e78790b0d843f834cc08eee455d1eed8d46c6
--- /dev/null
+++ b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9660f490aee5ed043bafb4f126a2b075b1c5efce5a939009d5aa721b0b5e3d6a
+size 424523
diff --git a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_zuco1.parquet b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_zuco1.parquet
index 6ea9a9b29e6baa2e971dbaaecc59b29bf245b9ba..d30efcd5221339d251113e299f8444c532806dfc 100644
--- a/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_zuco1.parquet
+++ b/lens_layers/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ceb728b050ed2e7eed2f24b4f776440f395ee57dbc3e41caffaa1816ea2c8217
-size 1064062
+oid sha256:f3bdd54a9815bbdce2873de30c3ad51dd17b370044304c237faadd1ca6145f8c
+size 1064186
diff --git a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_celer.parquet b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_celer.parquet
index d8dfc4831471a08373c530abcf7a5d45466e2516..420ee7bc99cf82a3ecbd70cfc79d451338bf538b 100644
--- a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_celer.parquet
+++ b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:de0e8982e0979af17feb7fa233e83a82eae8721208ad8314fb3feb163bf58013
-size 11274121
+oid sha256:6f66225e1f36573be56e3190413344023142f47961d902fb0df46c64f8d2e151
+size 11316536
diff --git a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_dundee.parquet b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_dundee.parquet
index d21d58b78b741d9cf1768a5c10f4b875913f536e..cd011eb8ac35ce6add79bc45a056be5a602de6d8 100644
--- a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_dundee.parquet
+++ b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:28450b5cd053a46924ac205411277789f9e744312d358614205cf18ae45bed4b
-size 2089048
+oid sha256:8af9f2d24fd1ec20d3e44a4b159dc887da5d9c88fe2da4fa4ebba26a6124212d
+size 2123187
diff --git a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_geco.parquet b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_geco.parquet
index 579a552914a876cd0a88c8fc0dfd01e6a9e39d72..8c5f752190dbc9a2045dcff82c8180ff79353af3 100644
--- a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_geco.parquet
+++ b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b8dd36eb3d173fab5a2ff2a9f0d24345a85a3f3fd0cd837561a941201051689c
-size 2222251
+oid sha256:e90bc3ac4defd51096150f4d2d84f7901222c2394b6f557b48e9d0b7fb28002f
+size 2221969
diff --git a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_onestop.parquet b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_onestop.parquet
index 25ec834e14447b593635e31192b3a8f4a55f9546..32300eefd0147f46cef9d89b08a77e8f65dbd109 100644
--- a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_onestop.parquet
+++ b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:565ea5f931f6f7604f3eb80dd01190f67101db79bece3ab6463e26dca5e1cf07
-size 1474295
+oid sha256:d5834cc518375b27c728bab3b999f4b87cc01873a54d0d23c1d9f296bdaec499
+size 1490561
diff --git a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_provo.parquet b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_provo.parquet
index c7ac654d5b93783e22e98b75e59d2568894a4698..8f1d0cecfb73235d9172174f51672275f965ba5e 100644
--- a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_provo.parquet
+++ b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:926605bf51270cfffbfde88d26f4f22e592ac0297994ad7e8fea9200ca49b0a5
-size 128190
+oid sha256:98aab487654927ff5a0d25b91d175d72d1265cb63dbc5f4007120711e7dcfe66
+size 132392
diff --git a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_sbsat.parquet b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_sbsat.parquet
index f4825b83a3d0b5bae8fa78f458b150ec9c91d5b6..0a02181a3b7452177ad620498131b6a7990b6f24 100644
--- a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_sbsat.parquet
+++ b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3aa76896058dc16a44218dfc0570cde10ca161dad3cd49eae49316fd76f47fda
-size 117235
+oid sha256:84bef6f74ef04534a8cffda6d4f5e6f25b19d3d7dab35fefd6019c055c5c07d8
+size 119024
diff --git a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_story_sbsat.parquet b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..7677b5c99876be0c81c6e0985044cd6694eacc88
--- /dev/null
+++ b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d894b79f9960c6103c7343878894476927f0e95ed793cbbdfd8365b66f25670c
+size 119326
diff --git a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_zuco1.parquet b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_zuco1.parquet
index f245a2797d782536f680dc060fd224fbce667049..6191017bb097bbbd262c980d13bcda87a727ef3e 100644
--- a/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_zuco1.parquet
+++ b/lens_layers/EleutherAI_pythia-14m_logit_lens_surprisal_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4a557dc5b31e0699be2dc7fb01709f2cbf28c0b46e58e48b9ac8498421d68a28
-size 322656
+oid sha256:5446eed65878d3bcfa06cf3b11837d93eaefa583760aeea5b1fa514337d292dd
+size 328920
diff --git a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_celer.parquet b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_celer.parquet
index 9c7fbc8d406ea9be58bf5613c63ef058ceaa53c9..ce7ec6c594b7abbe18bfc746c551fe45e813c53d 100644
--- a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_celer.parquet
+++ b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:92cf17e1691b0048ae31018d0440d7f91ac1bf7aec110a46bcb875791ea05780
-size 20028454
+oid sha256:676813be65037b529d85bc257aa66a1cf2ecc3cb75b8a1c85a8232eea0c0edf8
+size 20125843
diff --git a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_dundee.parquet b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_dundee.parquet
index 3a86222ee5c117f91fca0fbb937c3113f1dffb06..2489269a8fdd5d3a309808a88d77acce34776481 100644
--- a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_dundee.parquet
+++ b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:109b018d5ac303b7a6c23d3aab07ed61af717c0c9d2736886796c200f1fdcf70
-size 3538025
+oid sha256:cb85ee5abed984ea442a64eafc4858c2fdc456f8934a987d8e133d92cf19fa6d
+size 3592645
diff --git a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_geco.parquet b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_geco.parquet
index 60724ce672e6f22698db14a1415b9509dbd61dba..6ef151347a41182811a3c56c3bce96b148866b74 100644
--- a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_geco.parquet
+++ b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8109b7b5e25fe78c9d6ba75514094b22ca8939ac13237b60664a187c35b16809
-size 3808567
+oid sha256:1358e358f5053905a5321f5891c0568bebf8a2a76beb2b4fd991e8e1a9254b30
+size 3849441
diff --git a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_onestop.parquet b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_onestop.parquet
index 50006fd7ea3ed03e96dab63a8abe436746f0441b..55ffda49b2e3f95bf1f4ed932fafdcf5b21943a5 100644
--- a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_onestop.parquet
+++ b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b580073eeb16c5af682da09068af378813367a9dd885adcc4c2c1dac14f09078
-size 2521427
+oid sha256:deebf003a465cc04c1924b207b4168f1635b09694fe995aac1e55def37dcbbef
+size 2530071
diff --git a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_provo.parquet b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_provo.parquet
index 9c497a535e42075d1a041877be8295f78f62d219..f80d7a9eccfb46601c5c2a0f286565f70139001d 100644
--- a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_provo.parquet
+++ b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3c9f057ae3b31faad87cff0fa7574a7979ad6361d77d66de91836bb762d1831a
-size 249214
+oid sha256:cc3e7deb50fc474066a19512c1ab60a94bf29761b50c390d01bc3c1b841a9112
+size 249033
diff --git a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_sbsat.parquet b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_sbsat.parquet
index 5863477cb719168e69b7b8d1cb31a27fa37db376..7ec87c64391c9de7a78f733e3081bec0a927296d 100644
--- a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_sbsat.parquet
+++ b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4be6fe90ff2f1d1a41ecbfbce5f5c3161acb4636c29b5fc1c24a9127b71d53c5
-size 208553
+oid sha256:513a5395cf138613ad17f2efab6c73aed0333c8aea984e6a4803b59116630671
+size 211809
diff --git a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_story_sbsat.parquet b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..5c0c9af79d534c7a8a52ca725f71909e8fc62343
--- /dev/null
+++ b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fb7297eff79721b1a24a92258edeb881db25eb0d23dd8c0734637fdfaa5255c
+size 212209
diff --git a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_zuco1.parquet b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_zuco1.parquet
index 6eb51b0bbc6ad85637c089e1cd8be14705741d7f..da94311070fab4306d2f749e787b402bc59f04f9 100644
--- a/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_zuco1.parquet
+++ b/lens_layers/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e93073b96849bd15847229b7a8c31b3846758150d8f7e59bd5ef937b15a0edff
-size 597709
+oid sha256:490f501dcb8756a5dd4e23b496a73bfbe231dfe5c695d3f5bad35a9b23f8cb2e
+size 600256
diff --git a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_celer.parquet b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_celer.parquet
index 1a5c455f94a13bdef24105bb3ae20ae73a319f50..25e7d8fad2e3877bbcce189c5d9d285be2add38a 100644
--- a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_celer.parquet
+++ b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bc8885eadf9f8d07b4b1003503fd7d1b86aacf2603bc045311a12607b92b32ad
-size 25773772
+oid sha256:65c50ed7c91307e11b53cb550a930eba273a556897ad134601ccf6df30009d6a
+size 25867588
diff --git a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_dundee.parquet b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_dundee.parquet
index 2737a186b77d8813b855f89b8b675ef538cb782f..4ff5cca290bf4419a37fe9c6358a13b937b8fa16 100644
--- a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_dundee.parquet
+++ b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ba09cf4efc66646016823bbd2a9fa3c37bed8a0f6aba55412625cd37527803ea
-size 4506739
+oid sha256:d0053b538e6ef821e5dc02884384cf024ef15c959d76bfddb4021c95047065ad
+size 4589580
diff --git a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_geco.parquet b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_geco.parquet
index 26ea7eac30ce5c2ffd6eda9589fd51b6744a17d1..7afa58d83b22375e098cffbe31a526c8ef0ec2a8 100644
--- a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_geco.parquet
+++ b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:493801dd1b20692ee1a27cc7060cde2968680772df3e385a57e627c54040dcda
-size 4935901
+oid sha256:4592e86077f03e6626fd9b828e5037e9f061dd3e0849044ad6daf783a8cd53de
+size 4936791
diff --git a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_onestop.parquet b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_onestop.parquet
index d39f0fa4417ef905d1d0b91788c9bf273fc417fd..c3829e569129b04e3e70c097df46338fd8f4a90c 100644
--- a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_onestop.parquet
+++ b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9c38f7b69f34b074a12af1b1bd190b3e30632bd242550990e1108f29a90fec28
-size 3199890
+oid sha256:ede19595888e18552e4e151614b491071d3267d7579f3a0cb1d76fcefffbd55c
+size 3214361
diff --git a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_provo.parquet b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_provo.parquet
index 81f5d52dd4002cde64ac6c429063c6c544df63f6..dfff459e58ec5b34063d2d473964a66efcc686f0 100644
--- a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_provo.parquet
+++ b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:be0ffb7db805dc05d173ae1e6589f23f76e59f670c164e9bf15632ce2e48d782
-size 317444
+oid sha256:990fb5aa145ce6e1bb3a057878d900c7f49f4ec712b53424beea4cc749f9927d
+size 322975
diff --git a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_sbsat.parquet b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_sbsat.parquet
index 441ab8990c174b23aa6dd1aab43ea7af800264e0..2bb66fa41590c635eb7b3519f4c31cdc57601c7a 100644
--- a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_sbsat.parquet
+++ b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:81ef0ccddd45c461d7775925731146924a0031de591edc7a2ac96081dd8245c0
-size 292121
+oid sha256:4fb9a3103c20817e6e9d4cbf19b36ba95da0174def23531b71e1109e71242556
+size 290967
diff --git a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_story_sbsat.parquet b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..64f25f799798e0616ab9ca7bcaceb155a2e87641
--- /dev/null
+++ b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc26e6846e46c291e7b47c3060800481ca16d4446593b1e978f4b57d61dd1643
+size 292816
diff --git a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_zuco1.parquet b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_zuco1.parquet
index 36d6dbdf2771e473485cc183302051357880c558..6af93ca0319470be727044d8522d4630abbb9aa3 100644
--- a/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_zuco1.parquet
+++ b/lens_layers/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5c6fdb5db89363bf6e22aca753f7f4b81293143aa03e298922b3e4d6c719f66d
-size 768517
+oid sha256:17e70ae6f357347bfc1152e3b2066d0074f5267c6cfc430abec6e924d4be1c7a
+size 768722
diff --git a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_celer.parquet b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_celer.parquet
index 60a4ad5eb097199f5086c15b95b6980c905cb8f9..22e56d581633edcc655f4314182a239ccd8588f5 100644
--- a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_celer.parquet
+++ b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:724eb6134360d87f117763bff0d518b9c623cd59879e717a7d9f7059c9a8199b
-size 11378881
+oid sha256:11998e6d7570a77f3e3fea49851b5cee33cf1eee303f31f9922483f11c4eeace
+size 11416041
diff --git a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_dundee.parquet b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_dundee.parquet
index 9985a1ebee0acd985a8d8574f383a10cb29ff897..f370799a613ce2e782401360e8817d43288fe7d5 100644
--- a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_dundee.parquet
+++ b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:27014c9ee95054ea3556ded6f9ed0fe81ead470761bbb38e725a5988afa43500
-size 2110389
+oid sha256:c7b76944f769372f93648510c1a71c86c569c396d5c294e4cbf4fe3aea743620
+size 2138711
diff --git a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_geco.parquet b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_geco.parquet
index 17f4974db252e825baa7be8b74c3a9b5fcd6fa31..402adefec80cf884ff785e3b2b174a9a4ecbf34c 100644
--- a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_geco.parquet
+++ b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0f0be992fb64c03f3f536cafe34314d56a7c391d9996a3208415185ef5d575f8
-size 2235163
+oid sha256:89b229c1bd062ad1097693f9b074024a2e5dc36b7c5eb44f2fe3015dbaea3311
+size 2236233
diff --git a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_onestop.parquet b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_onestop.parquet
index 486040f276c260979f24ff1e9d2b78cf19d6af86..3355dc896bcfcaae38f0ca2b9f5f54f08aef2625 100644
--- a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_onestop.parquet
+++ b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3350b6f9f609fc41158ef8bad02701add94c2ab12054857dae7f749ebe3ec17d
-size 1492765
+oid sha256:305fcc6e2b830018fe064b72b70bb25356eab50d08903d116ac12c0d5029e387
+size 1502363
diff --git a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_provo.parquet b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_provo.parquet
index 7b2e3c59cb1f17f05b220d0ddc05a9d9d2334edc..a8e84c610c956f22acf6f511bcfffe5f3eb4b7cb 100644
--- a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_provo.parquet
+++ b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bc8e69dcd2adc3631b302d482a5119b1bd43cad05b80aa44fc04d57d4d054930
-size 128734
+oid sha256:109d8e7da62bb983e09aebed2e8a38c781848932a50e6903df6e5662939e27b8
+size 132980
diff --git a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_sbsat.parquet b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_sbsat.parquet
index d584b00e70c5d8fac165ce1120ed6e436b9073de..f5fde8e9b5192464a498f78bd26c427d30a93979 100644
--- a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_sbsat.parquet
+++ b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4c81d056fc115567e835d92a68791bf56c03ec1e5a74c8edfca51bf9f3065a2c
-size 117969
+oid sha256:35a8c409ffa36c753d0bb48143701c5f3e5233ff83dce7d2eb108121bd2ac679
+size 119586
diff --git a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_story_sbsat.parquet b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..d65484c3f0809c24fa4937007548afa483b3fc6c
--- /dev/null
+++ b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:200b4f5136a1744677aac7a4e88692fe8b8937b3a1fe3f0ce99b98ee88dbf84d
+size 119885
diff --git a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_zuco1.parquet b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_zuco1.parquet
index 71265a58823f1a7fc6e1bf037280d27a7578b7df..71ffe604ebe02968c33f1458c28862c0a9cf9faa 100644
--- a/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_zuco1.parquet
+++ b/lens_layers/EleutherAI_pythia-31m_logit_lens_surprisal_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b0213a3701598a881df101e6b58b25ecce03d51243a46b688f77432d917c0edb
-size 325998
+oid sha256:3ea8d952980a4b508c5e895be75680213355e12bcb2226b5a907a6c901920177
+size 325295
diff --git a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_celer.parquet b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_celer.parquet
index ceed650e73b1e718f655b21d8baa93f44a6276e2..7bad8988428ce3cf239210832e9c1cd9553547e6 100644
--- a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_celer.parquet
+++ b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4d0b81da4ccb6ed220d033dda10bd4425d1a8644c18b278f00142c95836d76eb
-size 37268931
+oid sha256:b4dead6a138a0f4fa1794edd68fa6704b1344ea74fbbd92ba9074e905551725d
+size 37347699
diff --git a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_dundee.parquet b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_dundee.parquet
index d05e5432ecf9ed1ab136e69efc61af828c2d8606..c14b661d8bfc6b5f5f8607dc1dff15e2d460bad2 100644
--- a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_dundee.parquet
+++ b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8b583530cd8e3b33e6a7d9fbd5d1c053f090f6afc7cea419103f3b084ff8e4e3
-size 6366895
+oid sha256:d34b699b1445471cac651c05ed9271cc285050035744e040cafa2186117e5534
+size 6476332
diff --git a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_geco.parquet b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_geco.parquet
index 931a4bb7ef3be4ea7893ed5ab5e7542aed5663bc..b4101f0dd260e7527d444b98a4552cfc537f1179 100644
--- a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_geco.parquet
+++ b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4026db13de95abd1c85f5833995f86575c88b1a8bb3f3434424258e9dc69f2c6
-size 6986607
+oid sha256:bf36d008e23b633a37172ce9fb0752329e21122748fcc196f7da29c6676aef91
+size 7008170
diff --git a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_onestop.parquet b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_onestop.parquet
index 64c1c751aa1521ccc60d6f8f63bbf69ca3819f37..a53ab3f33433133d7ea907ce61a15a08b67cef84 100644
--- a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_onestop.parquet
+++ b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e0b3aeb928040526d749ca496c0b2c649c8c882f2e4e2c498bd61d74652383d5
-size 4539714
+oid sha256:b6158c3891e0c09ebd164bc758011da92178d9a134b053a75794f1da102069cf
+size 4540149
diff --git a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_provo.parquet b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_provo.parquet
index 9c2be7bf58e2673672e688d60d29ec523ed6c8da..d67dc56abed4276758f24dd7dda69e4fcd058dfe 100644
--- a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_provo.parquet
+++ b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d7f88a117a5378c510fb35c8287734c5c5833eaa662c777f090952051e200ec6
-size 466590
+oid sha256:f32b30e965b5876408898dca4fc1524ad67739baf6df001ba2eb2291e49dc590
+size 476069
diff --git a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_sbsat.parquet b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_sbsat.parquet
index 9deedf19e0f2c0e2412b584d196c76ee1c9ae16e..4cf25c473559803036cd29d2ad637ed9b2a8872c 100644
--- a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_sbsat.parquet
+++ b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c58772639dd09f6fd5ce1a94cafd235727487a78abe3cdaf995db810b3492e88
-size 423939
+oid sha256:61ee2acb5658c30f57920df49802cbaf7bd7e411506c21f542fc106d19014aec
+size 422038
diff --git a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_story_sbsat.parquet b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..e39e6a4a14fa67b4eb0908d5f308b83120924de7
--- /dev/null
+++ b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec12fd1639eef228f82fd9a7d97c4940ba994861abc9c6263c21cce66b111ea9
+size 423842
diff --git a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_zuco1.parquet b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_zuco1.parquet
index dff4ad7c1a16b0d31250266e2d17630f44be7981..1895cad8694e5fa6fcfb3146a096599e1992ca5d 100644
--- a/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_zuco1.parquet
+++ b/lens_layers/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6a34869f3448a2799611ff0cbe2e3fbec522175729291b69e1170ae184bd1d3d
-size 1062706
+oid sha256:1ba69c9afc63d8e8b6708a68ea49f8e07137f75239a482508ea81cc020eee546
+size 1062987
diff --git a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_celer.parquet b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_celer.parquet
index ee0fa96003ffa4e92f477f08bc47f5d0b709067f..f21d423ac64d91cbebba1d6c6081d89cdf3320fd 100644
--- a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_celer.parquet
+++ b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:26f1a22c0d53d1224cd82b25f0ed8c252da0a6eeb66bf9cf7ba3ce284e97fe61
-size 11402345
+oid sha256:9f974bec3467bb7c7b0bdaba5a6135a4492f9f7f4b6ad234e43fd48b5e3a98e9
+size 11473781
diff --git a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_dundee.parquet b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_dundee.parquet
index c6cd09ee95ed1cd11e17b409130ba4b4e44157f0..d6e834b839039b47dc6bf543fcfff81fafcc6591 100644
--- a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_dundee.parquet
+++ b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0ee016ce45983d66e942cde95a514601e5c4697073f0c91ecf67931d249fbf16
-size 2109181
+oid sha256:f4e7c6aed91042275a1184158f7e557dab3d9cf34796598e11c5490baea14fee
+size 2140319
diff --git a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_geco.parquet b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_geco.parquet
index 3e5129d2d534ae8f516a5b0e95b70fc7a590a324..20eb308b4010f3559045615c5cf02c350708d061 100644
--- a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_geco.parquet
+++ b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fd8c4c6a1d6dbe1919d85581761311830a9090f49783d426550e042ada803ed0
-size 2229339
+oid sha256:c42585568439a71eefab3290a2a68d5eec0c562870d4501ce12b5d237bc5900a
+size 2251071
diff --git a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_onestop.parquet b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_onestop.parquet
index a63492f75ed6637d3ba3ce62a25cbc1658a9d6f0..fe796f3d8d447fdc0b9241f030d6b47ef908dd78 100644
--- a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_onestop.parquet
+++ b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:864134cca2496638581b209de6739b9223a1939b15d3f8aed1e9f6847058ea29
-size 1494160
+oid sha256:1f877fef27a1e95340cb3285e015f343cf817b7f164474a3ba701c195d7b9eb7
+size 1511089
diff --git a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_provo.parquet b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_provo.parquet
index 51f073f66ca5d31fe4d473c61d23e9a720fb0243..ee46452ee2b125c17d6c366a3293b16474d71876 100644
--- a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_provo.parquet
+++ b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e5b4c804abf22772409d1adf4266c58b0e35d84f8647dee2c00fe52be8a68e66
-size 128747
+oid sha256:221a15728e68dcbaf233475a6123547e4ea5c6a5823a8612bbdd3b02bfd36d57
+size 133302
diff --git a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_sbsat.parquet b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_sbsat.parquet
index a16269cf56d124085d8a1d7bc44bbd5cf01a85a8..40c715d83e6be011e610e16ce0117ee5365060d7 100644
--- a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_sbsat.parquet
+++ b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:74bd3ad2e37e8a44d1c6b792f322eeec58baa6510aab673114cc97b78a9d7917
-size 118014
+oid sha256:5f9b69107a5349a0e62262acdc242913b853f45af968d3aa289f276b58a9caab
+size 119846
diff --git a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_story_sbsat.parquet b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..915346cf5419d8536c67ca93385a116c700e6902
--- /dev/null
+++ b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:771c5dae169f3df6e38a6c0945bdc1c65ef25fb342db5a56e6daa9c5b1fa0303
+size 120447
diff --git a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_zuco1.parquet b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_zuco1.parquet
index 4fc7b8440fdab70f45880e27249d04f43761ee07..a088c9f6e2a2ba6b3707e9dc4a7672012825aab8 100644
--- a/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_zuco1.parquet
+++ b/lens_layers/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:05c1b829e350a396b8020e9798500f3283d56bf9e239de25f86e1d4123e29451
-size 330422
+oid sha256:cccf0fc82fce944615d534923674a55f36038aa61e0c7cbdf7c4f504bacd1df5
+size 327301
diff --git a/lens_layers/gpt2-large_logit_lens_surprisal_celer.parquet b/lens_layers/gpt2-large_logit_lens_surprisal_celer.parquet
index 72d4c82711562611a21c6a8a5d29fedc90fe47f2..4039cad6f002080d49eec849fe72a67db8a2d773 100644
--- a/lens_layers/gpt2-large_logit_lens_surprisal_celer.parquet
+++ b/lens_layers/gpt2-large_logit_lens_surprisal_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2a3f7fcc10fb5cd4e64db5ae1d8260e836076859cbac12a0e4b84fafa9e5da17
-size 54443115
+oid sha256:31557c113390a8713bea8579a25f710a6386aa485a3062ac556424a01ac64206
+size 54819439
diff --git a/lens_layers/gpt2-large_logit_lens_surprisal_dundee.parquet b/lens_layers/gpt2-large_logit_lens_surprisal_dundee.parquet
index 5b1e3ca67cd75b460f93546d40fa791701ee3082..3138a11cba38d94a91bf24fe13525b3f77fcee31 100644
--- a/lens_layers/gpt2-large_logit_lens_surprisal_dundee.parquet
+++ b/lens_layers/gpt2-large_logit_lens_surprisal_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d59e12defec0a7885cd9c51266a9270d61d1d806b596332ea02ab80d4521ff3f
-size 9242629
+oid sha256:ddd91c1bae64c7c44646339bde94aa81886490135cb7b0e738bc538b2a5a568b
+size 9462720
diff --git a/lens_layers/gpt2-large_logit_lens_surprisal_geco.parquet b/lens_layers/gpt2-large_logit_lens_surprisal_geco.parquet
index e33160f8c4f49031d67abc6a8cbec26910150d0a..6803fc06acb0af4f3c23e473471e2a2a7ca5888c 100644
--- a/lens_layers/gpt2-large_logit_lens_surprisal_geco.parquet
+++ b/lens_layers/gpt2-large_logit_lens_surprisal_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:858291a97193811702b716d8c32a19992458d28d89fd41722bbaae66c0b19c76
-size 10122068
+oid sha256:5ba0049ee13b69523490e87e526a790a42b4c75580d37edf88bcf3269b824d96
+size 10188754
diff --git a/lens_layers/gpt2-large_logit_lens_surprisal_onestop.parquet b/lens_layers/gpt2-large_logit_lens_surprisal_onestop.parquet
index b1cc97d808895bccb8a4a25147ea551cf893e1cf..f700aba4069ae54dc2035b17319c8291fe79da0e 100644
--- a/lens_layers/gpt2-large_logit_lens_surprisal_onestop.parquet
+++ b/lens_layers/gpt2-large_logit_lens_surprisal_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4bd5acfeed11aab35826ba4d1d019f4a5b191b94f37f5e96a537603ee84affb8
-size 6578269
+oid sha256:e14d6e651f02abb70ca2b1688255d99a8daacd7980041487b3ad3309678e7d1a
+size 6644642
diff --git a/lens_layers/gpt2-large_logit_lens_surprisal_provo.parquet b/lens_layers/gpt2-large_logit_lens_surprisal_provo.parquet
index 72ca5b86d47789fe4f931d8525e775fb045ab95e..41727405cdc2f71803328e37e3382c6afc5e743d 100644
--- a/lens_layers/gpt2-large_logit_lens_surprisal_provo.parquet
+++ b/lens_layers/gpt2-large_logit_lens_surprisal_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:39fe78be41fe8842af55a1d26d753e282c9a0d61892bd61249159bc228afa05c
-size 683954
+oid sha256:28860b67635757cdd666504ebac9d3c9ef1d8111eb9f1892f370df1610712204
+size 698227
diff --git a/lens_layers/gpt2-large_logit_lens_surprisal_sbsat.parquet b/lens_layers/gpt2-large_logit_lens_surprisal_sbsat.parquet
index c09db570d924e63e5ec17b0ced973cad4fa18832..abbf8bf1c1496357086b831a08293c91ab2c67de 100644
--- a/lens_layers/gpt2-large_logit_lens_surprisal_sbsat.parquet
+++ b/lens_layers/gpt2-large_logit_lens_surprisal_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2c0b17dff7e8e21497a9705b6962bcd82e1ac5ed0494f6c4abefde69ef174c80
-size 631092
+oid sha256:4ac8531b3aeae5e943760f1f1d442e093ef71b1a04affc3ae2e36a04d4d1de0f
+size 632041
diff --git a/lens_layers/gpt2-large_logit_lens_surprisal_story_sbsat.parquet b/lens_layers/gpt2-large_logit_lens_surprisal_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..2080e91f4ede552803e0316e77ccc108121f53e3
--- /dev/null
+++ b/lens_layers/gpt2-large_logit_lens_surprisal_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:053f3208de228d486b573325a4bf4b7e883e4529f6a2221fadc2670ee22df71f
+size 635645
diff --git a/lens_layers/gpt2-large_logit_lens_surprisal_zuco1.parquet b/lens_layers/gpt2-large_logit_lens_surprisal_zuco1.parquet
index 648db30345305f3fb63cf67c54b53b3e714b5d3b..3b65ba23493ef4876e49e051411691c006a7339a 100644
--- a/lens_layers/gpt2-large_logit_lens_surprisal_zuco1.parquet
+++ b/lens_layers/gpt2-large_logit_lens_surprisal_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c6f6b79a8087080a16adc0c1e76170b7fae40f506f00abe04d14a608d921d0bd
-size 1434917
+oid sha256:2f45523f2f40d97faf96c533064f607301799f5d6446ac9a0e1882fa49a91d07
+size 1437376
diff --git a/lens_layers/gpt2-medium_logit_lens_surprisal_celer.parquet b/lens_layers/gpt2-medium_logit_lens_surprisal_celer.parquet
index b68ca8c1c7a672c77801f01ac47ca54c0c998884..a0ccaacd99e20bcfd03c1ebcca6a545b2d882679 100644
--- a/lens_layers/gpt2-medium_logit_lens_surprisal_celer.parquet
+++ b/lens_layers/gpt2-medium_logit_lens_surprisal_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ec897be99eb5056564fd1174c824d31a379f11d04cef6286e198a3dd1d9185ac
-size 37759864
+oid sha256:b98dce54e9b33ec0246718e4ef9b15c7ed8f5119fdd788ebd16a79347cf90d22
+size 38143823
diff --git a/lens_layers/gpt2-medium_logit_lens_surprisal_dundee.parquet b/lens_layers/gpt2-medium_logit_lens_surprisal_dundee.parquet
index 069e752feb2526f4df5278824a167d7672b1627e..2577011e249c44208c849b79cd21737627695519 100644
--- a/lens_layers/gpt2-medium_logit_lens_surprisal_dundee.parquet
+++ b/lens_layers/gpt2-medium_logit_lens_surprisal_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b917ffebb75fdb572adb142121976f9367fe412fe59e67a2fb6d15725e3e9e4b
-size 6497898
+oid sha256:e272fd2c5681a575df564c1a21fffab044d022de58f170993ff26d473c2883cf
+size 6616264
diff --git a/lens_layers/gpt2-medium_logit_lens_surprisal_geco.parquet b/lens_layers/gpt2-medium_logit_lens_surprisal_geco.parquet
index bad166c7fe6bedd0e7d257d1fb3bdabc735e2e20..abf202dd63c8df998b638c98311c427f8c8745fc 100644
--- a/lens_layers/gpt2-medium_logit_lens_surprisal_geco.parquet
+++ b/lens_layers/gpt2-medium_logit_lens_surprisal_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f770ca92eb472eadd716d4292b193f8b56bbee51ad3107e4e03cbc9c63e6d54f
-size 7039405
+oid sha256:727ec1a55a535214f39c055ad4bf9284ae56cdfb92b8415bb234be3f682f1e3e
+size 7118872
diff --git a/lens_layers/gpt2-medium_logit_lens_surprisal_onestop.parquet b/lens_layers/gpt2-medium_logit_lens_surprisal_onestop.parquet
index 6a5dee1d34492bba5fce967855ba9d2db232b016..d26b67efc7d7adf8eb1d499c71e8ce6a234baf4d 100644
--- a/lens_layers/gpt2-medium_logit_lens_surprisal_onestop.parquet
+++ b/lens_layers/gpt2-medium_logit_lens_surprisal_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6d2a25ceb26c23b4cc10cdd94185c60220fde409d1638c2247ce96e71161a238
-size 4613783
+oid sha256:5b595fe2d2b0ec203e473ce751eb0201f42427c525c4e9c3392ff78771606003
+size 4664987
diff --git a/lens_layers/gpt2-medium_logit_lens_surprisal_provo.parquet b/lens_layers/gpt2-medium_logit_lens_surprisal_provo.parquet
index 5551541ef8df4ed944e019f6d1874977662373ac..093501d663e7bcbe3df3a8e76c06fe188459e050 100644
--- a/lens_layers/gpt2-medium_logit_lens_surprisal_provo.parquet
+++ b/lens_layers/gpt2-medium_logit_lens_surprisal_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ab7fb78e1449214dbc7679dda1ad2746bc4b6a7f2013a327571fa1f42324f8ea
-size 473483
+oid sha256:b2d12e082f7852f130ecd6764612a4397de5d5b6f60e931f9a0847dc2d73eade
+size 481866
diff --git a/lens_layers/gpt2-medium_logit_lens_surprisal_sbsat.parquet b/lens_layers/gpt2-medium_logit_lens_surprisal_sbsat.parquet
index 7cc99c1aa58ecee40a969028ffda3586f079f65e..1c32ed371c9598bd6c954a62a655ecf5b0a17b3b 100644
--- a/lens_layers/gpt2-medium_logit_lens_surprisal_sbsat.parquet
+++ b/lens_layers/gpt2-medium_logit_lens_surprisal_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ee077c4f424380aa6a70d6596ddf3e95ba93846723bea66f108584e1457a25bc
-size 424584
+oid sha256:ebf1c55b17fe68e4f974cab5536745fc2e1744857dc3d67d36e0ad6fd6908229
+size 426144
diff --git a/lens_layers/gpt2-medium_logit_lens_surprisal_story_sbsat.parquet b/lens_layers/gpt2-medium_logit_lens_surprisal_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..0aa7d41fb428782a670234bb6b034e791d67258e
--- /dev/null
+++ b/lens_layers/gpt2-medium_logit_lens_surprisal_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b1aedea5c2078ad14e44120d78e7d15b17c9c261662f0c5f0978b847984c52f
+size 428642
diff --git a/lens_layers/gpt2-medium_logit_lens_surprisal_zuco1.parquet b/lens_layers/gpt2-medium_logit_lens_surprisal_zuco1.parquet
index 2dfaa12fcc1344d7b7705dc32e9961061804b6bf..6a9a4e938b6ef35f0f4fbc516e382994ce44b948 100644
--- a/lens_layers/gpt2-medium_logit_lens_surprisal_zuco1.parquet
+++ b/lens_layers/gpt2-medium_logit_lens_surprisal_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1e2c3f2c3bc4eb75aed71aee76f3b842449af346e6fe39080a894d2bbfe1d49d
-size 1069605
+oid sha256:17accec0115b08ed855d1fc7399cafec928a97b08395a8212b5c41a202316e79
+size 1079497
diff --git a/lens_layers/gpt2-xl_logit_lens_surprisal_celer.parquet b/lens_layers/gpt2-xl_logit_lens_surprisal_celer.parquet
index 1fa0d6759a38ff44b1502f52da2c016528e5dd4d..be43e1b683b61b6aa6b7331f76b73b900c2bc416 100644
--- a/lens_layers/gpt2-xl_logit_lens_surprisal_celer.parquet
+++ b/lens_layers/gpt2-xl_logit_lens_surprisal_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5aced1921e34c302cc6285587cae88d0443f8cc39d71810e29e14a76deae7aca
-size 71135171
+oid sha256:ee1bac8fc8801a0c5557ac0501da3dd21e30f0698d7d56348b122b7bc11a6f62
+size 71686144
diff --git a/lens_layers/gpt2-xl_logit_lens_surprisal_dundee.parquet b/lens_layers/gpt2-xl_logit_lens_surprisal_dundee.parquet
index d7d5b32cf52857dd2973934fd507d6c9da4e1183..37fdc7075b16bc49cc5e5134c9d3a48d34517366 100644
--- a/lens_layers/gpt2-xl_logit_lens_surprisal_dundee.parquet
+++ b/lens_layers/gpt2-xl_logit_lens_surprisal_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a71cba49657052befd7e5dacd3bcc424dfe7f10538abfc102efa71245068bc6e
-size 12019049
+oid sha256:ebaeff7a612e8532a73020e0d19e9ce8445b57d47bdc4f683d7b531bc3b108b2
+size 12234228
diff --git a/lens_layers/gpt2-xl_logit_lens_surprisal_geco.parquet b/lens_layers/gpt2-xl_logit_lens_surprisal_geco.parquet
index f0f74fbeefdc6a0eeba02e5887ef91f19cc4b7ea..06eeb13bd4b433f0c9867d5905a3fa8b86bf85b0 100644
--- a/lens_layers/gpt2-xl_logit_lens_surprisal_geco.parquet
+++ b/lens_layers/gpt2-xl_logit_lens_surprisal_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4c2f8cac9b01edf4f2f5f77fc6a71eda8f0bfa4f28fddd8f77d1dee32c6ac8bc
-size 13160433
+oid sha256:5569ddb184614f56d2b83db717b725dc71fece34fac299b6154812512c2ed0bc
+size 13278764
diff --git a/lens_layers/gpt2-xl_logit_lens_surprisal_onestop.parquet b/lens_layers/gpt2-xl_logit_lens_surprisal_onestop.parquet
index ca05ca3a6dbd07b1603e1041feb37b81a75a4073..bfb9d6d630dbe06c434f9426095d4b08d5474521 100644
--- a/lens_layers/gpt2-xl_logit_lens_surprisal_onestop.parquet
+++ b/lens_layers/gpt2-xl_logit_lens_surprisal_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bc13df3d259b6d6de7bf644df08555fafbb592a404705ddbdb3f1ca6da7ff3bd
-size 8521098
+oid sha256:611343a62fd2d0c5f0448a9297c9b6d25bd4f1576e004cfd99ee0667d48c9daf
+size 8660810
diff --git a/lens_layers/gpt2-xl_logit_lens_surprisal_provo.parquet b/lens_layers/gpt2-xl_logit_lens_surprisal_provo.parquet
index ca5e52db0a0308b368ab7bc247cabf17e63b90af..8e5e2a03231fe32c0bde3511c360bdad04985837 100644
--- a/lens_layers/gpt2-xl_logit_lens_surprisal_provo.parquet
+++ b/lens_layers/gpt2-xl_logit_lens_surprisal_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:53787880871d2e50c9aba6edf2dd557ada0b2c45cb0d1296df4fe5d3008a4c11
-size 893154
+oid sha256:1c99ffef5a7bd88fff2fff78ff20dd724f260bcbd6f2e525727c5c1ba23884c4
+size 924072
diff --git a/lens_layers/gpt2-xl_logit_lens_surprisal_sbsat.parquet b/lens_layers/gpt2-xl_logit_lens_surprisal_sbsat.parquet
index 7c0948ed44ba6dadc3e04277fc53b7bc98697c00..f742d92a32b3b8cfd34a1bc2531e12cd454ed75b 100644
--- a/lens_layers/gpt2-xl_logit_lens_surprisal_sbsat.parquet
+++ b/lens_layers/gpt2-xl_logit_lens_surprisal_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:56fd42f9c31ad2da573a6af742d3870fd5795ff39bc5725641af104b543d8b7a
-size 826161
+oid sha256:adf938c91f2ff2459288f2271be114fc76165a3cd64dd5ee4b271308affb59ea
+size 828574
diff --git a/lens_layers/gpt2-xl_logit_lens_surprisal_story_sbsat.parquet b/lens_layers/gpt2-xl_logit_lens_surprisal_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..4e2bb02dbb1ca932ceb0aec6caf96a481904b8d4
--- /dev/null
+++ b/lens_layers/gpt2-xl_logit_lens_surprisal_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:342fc24b7b59ea7bdd85ea212819cbcecab6eccc44fe44cd233acd335c84139c
+size 833368
diff --git a/lens_layers/gpt2-xl_logit_lens_surprisal_zuco1.parquet b/lens_layers/gpt2-xl_logit_lens_surprisal_zuco1.parquet
index 511f59d273eacf78f54174d2fe9537d88305bab2..6377a0281fa1998281e19c94ba2092f593b64177 100644
--- a/lens_layers/gpt2-xl_logit_lens_surprisal_zuco1.parquet
+++ b/lens_layers/gpt2-xl_logit_lens_surprisal_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:56fe1dc5ca4d92455da0e583d76dd91f3bbfa719fc7802b3d43e35a830a622cb
-size 1782905
+oid sha256:c7f167ec63291c71221d573bc98ea452fb52e1386c58faf652ebb426583070d8
+size 1791675
diff --git a/lens_layers/gpt2_logit_lens_surprisal_celer.parquet b/lens_layers/gpt2_logit_lens_surprisal_celer.parquet
index 1cc1e1952391af76508e2d6e22842f63ebd51978..c3c5d68eb1b6d942206b2aae6da925745251e382 100644
--- a/lens_layers/gpt2_logit_lens_surprisal_celer.parquet
+++ b/lens_layers/gpt2_logit_lens_surprisal_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:65a65bf236ffc991290b9de3f7daa9619e44b90c1a8e1f8ab728acbddb3c99b1
-size 20301819
+oid sha256:5077986a6cf1a52437501f4f03b47e53afc58a1d21b64fe80f4a471d80d6c320
+size 20549841
diff --git a/lens_layers/gpt2_logit_lens_surprisal_dundee.parquet b/lens_layers/gpt2_logit_lens_surprisal_dundee.parquet
index d6e5b304560549ec94671aeceadabbe08c200503..569eb048d4eea112ed489666f378738330a497eb 100644
--- a/lens_layers/gpt2_logit_lens_surprisal_dundee.parquet
+++ b/lens_layers/gpt2_logit_lens_surprisal_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:53276edccc235026c61dc8affe784f64c9a8823032b1ed86c841307192cd2901
-size 3616934
+oid sha256:5d0018870b9c4433dd434b26e7ddca13007bf75c109d5be92bc66ffd76389f7e
+size 3682008
diff --git a/lens_layers/gpt2_logit_lens_surprisal_geco.parquet b/lens_layers/gpt2_logit_lens_surprisal_geco.parquet
index 81010109d113cde7b89bbe1242eb1420f88e9b17..4389bb6f3f8cc37aff5da839c70b29996500655e 100644
--- a/lens_layers/gpt2_logit_lens_surprisal_geco.parquet
+++ b/lens_layers/gpt2_logit_lens_surprisal_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bc256739a5a4d104c5a8632f73488a8a263c989855e98998b75d9799ce5b412c
-size 3888121
+oid sha256:09d72855daec059b68ab9cee532d640ee2d3c9658deaa71e81205c39c1590342
+size 3922454
diff --git a/lens_layers/gpt2_logit_lens_surprisal_onestop.parquet b/lens_layers/gpt2_logit_lens_surprisal_onestop.parquet
index 1e689eafbb4c5abcf4fd233d3237b1f02927dd03..be7ecccb8744052154ca6863963e76f720363a26 100644
--- a/lens_layers/gpt2_logit_lens_surprisal_onestop.parquet
+++ b/lens_layers/gpt2_logit_lens_surprisal_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ae0cadd49637566808b90569ecaab5bf4a960a3635571e326628540fb7ab2b39
-size 2582066
+oid sha256:5cc7437b9b8b7729b3642957d05300f5cbfbc204086b2b60f40408570b5cd4b9
+size 2613277
diff --git a/lens_layers/gpt2_logit_lens_surprisal_provo.parquet b/lens_layers/gpt2_logit_lens_surprisal_provo.parquet
index 2330319f5c4a8ae8be2b4bf3756303b0507714f7..a9357d405093f494b5b0da4d6a713db398553191 100644
--- a/lens_layers/gpt2_logit_lens_surprisal_provo.parquet
+++ b/lens_layers/gpt2_logit_lens_surprisal_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a8f395cb610254a1c50269204ba575c8d748cafbcc7f348fb1bee5041de342b2
-size 247685
+oid sha256:3acc99dd3c2470934570beacae7597793295047d2b0cd68e86488c72a2c4796c
+size 255183
diff --git a/lens_layers/gpt2_logit_lens_surprisal_sbsat.parquet b/lens_layers/gpt2_logit_lens_surprisal_sbsat.parquet
index 7c3c9720d5fe9b9396c93518762d339bf0552526..6ba212637a0576a63e60ed59c91d3acbde1f4c06 100644
--- a/lens_layers/gpt2_logit_lens_surprisal_sbsat.parquet
+++ b/lens_layers/gpt2_logit_lens_surprisal_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:902e6a3d4baa77dd171b563c834e1dbb312e1cf9d60a244d403af4f3b005918b
-size 211155
+oid sha256:56ae52411fc24a8c1b9b25aa038b5f0c32c0e3bed717b512f764cd41576f8de3
+size 215423
diff --git a/lens_layers/gpt2_logit_lens_surprisal_story_sbsat.parquet b/lens_layers/gpt2_logit_lens_surprisal_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..9bb52835d7e49b7e4f856431715ba1da227ef022
--- /dev/null
+++ b/lens_layers/gpt2_logit_lens_surprisal_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac2537bfc2cb2f39f47b7a4476b9027ac373d260e27903cac8e62d43b2bf1488
+size 216449
diff --git a/lens_layers/gpt2_logit_lens_surprisal_zuco1.parquet b/lens_layers/gpt2_logit_lens_surprisal_zuco1.parquet
index 699b8a8ee9ee142a572e253889fbc350753fc255..0506ded6367de7007d3c7ee5e73d49675010decb 100644
--- a/lens_layers/gpt2_logit_lens_surprisal_zuco1.parquet
+++ b/lens_layers/gpt2_logit_lens_surprisal_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cca932582ef48e21af9c597b431e7d56ce5d88a99d8548e6ef0e4fa9189b56fb
-size 601563
+oid sha256:653ceeb9bce0ffb94a25dd6508cd8a2e0371f9b5f114760186decfc8db71999c
+size 606769
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_celer.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..abe9ee4e62bf429a1d44180ebe911af097248f90
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_celer.json
@@ -0,0 +1,130 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8f58ac0fa4fe531235ab96ad1c3a7462d95e35f8fd33cbb197a9489cf430f352",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:33:00+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_celer.pkl",
+    "source_sha256": "2e0c86cde8a20b045da488e6834e906b8f64274cf3c0988e8e826a83c695044f",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_dundee.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..322cb7967488886f012a418ecc6b6178a1d5c590
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_dundee.json
@@ -0,0 +1,130 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7f49c707a778ba1e7014550e32c3550935b6b6c8ce41bdc0401dcf710505c8bf",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:13+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_dundee.pkl",
+    "source_sha256": "75e1c6024c556225b88db5b403c1922b8d223883efce3bd74102bfcae4f0bfe5",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_geco.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5265f6c6674a5cc3fec3c2f31bb9d43353bd32f
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_geco.json
@@ -0,0 +1,130 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ba9275e81806002c58736cbc44cd50a66ab00a6a30f99bbedecedc57c8961a66",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:40:42+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_geco.pkl",
+    "source_sha256": "87d8121b308bf27911f83be6c2b57e729af1a5f4de3d3dd688cba7f1f59c0135",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_onestop.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..47b790e084fe7eecf3289657faf40b10a1b7657e
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_onestop.json
@@ -0,0 +1,130 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5b3ade99d9aa8a506f95a8b002eb90bfc1de53836f8da4daac354500b5d52e68",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:33:33+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_onestop.pkl",
+    "source_sha256": "45fc2c9e7dab87e965a542dcd900b0e6386fc4cb1a7d40a94b062dbf85034232",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_provo.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..0bb8ee5c239feff0b3e8b0d416a64f67a18632e1
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_provo.json
@@ -0,0 +1,130 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "046079c44e5efb55ddfe2febad8b93f450278b9431d5930bc7813cf4ddf6eff5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:37+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_provo.pkl",
+    "source_sha256": "28d8bb886b5dbce485249d8ab522a0ac8beb2323edb07aec5836fa61c5212a41",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_sbsat.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..190b107cd9d3fd4926bdf999249847ec77a64da4
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_sbsat.json
@@ -0,0 +1,130 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7f7fecc6d344b60875b6c21b69cec6a2de68102683a05b51907822ce100d9489",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:03+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "775722cb3e216909c93568a4acdce88e8ae19513f20c4802419a883e8d29af0f",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_story_sbsat.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..f97f7d58af31c82aed6babfbf2fab32743926583
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_story_sbsat.json
@@ -0,0 +1,130 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "40f9d89275dd6f6eb7b2911aff57404610342dc25dc0f220fcd1b5c5ad6da8dd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:48+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "9a77c4b4abc6e1b83b2defb829990a0b671e62ad12f914cb72dde5f5c2df305b",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_celer.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff6e129f76177d639f3c818958e46fa530a829f3
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_celer.json
@@ -0,0 +1,101 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2e0c86cde8a20b045da488e6834e906b8f64274cf3c0988e8e826a83c695044f",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:32:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_dundee.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..67fd7157538ed53e7d4afbe7fd8f9456e3273f0b
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_dundee.json
@@ -0,0 +1,101 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "75e1c6024c556225b88db5b403c1922b8d223883efce3bd74102bfcae4f0bfe5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:24:58+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_geco.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..3606e5dd22e68a0f876fda6ac7bc58b84f191440
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_geco.json
@@ -0,0 +1,101 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "87d8121b308bf27911f83be6c2b57e729af1a5f4de3d3dd688cba7f1f59c0135",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:40:23+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_onestop.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..ebd3b720566643551b8b94a5c30d2282ba352528
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_onestop.json
@@ -0,0 +1,101 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "45fc2c9e7dab87e965a542dcd900b0e6386fc4cb1a7d40a94b062dbf85034232",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:33:19+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_provo.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..023485a3d7bc38dde7ab06c6f1b50da7183bf577
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_provo.json
@@ -0,0 +1,101 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "28d8bb886b5dbce485249d8ab522a0ac8beb2323edb07aec5836fa61c5212a41",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_sbsat.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..55c036c20b6d401efecae27fca7ae1b5bd085cca
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_sbsat.json
@@ -0,0 +1,101 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "775722cb3e216909c93568a4acdce88e8ae19513f20c4802419a883e8d29af0f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:02+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_story_sbsat.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..41264021e8642c71b5087cf54b0da62cf227121e
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,101 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9a77c4b4abc6e1b83b2defb829990a0b671e62ad12f914cb72dde5f5c2df305b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:38+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_zuco1.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c132f34fea0997d89db849d4faed04ddc9ff097
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_zuco1.json
@@ -0,0 +1,101 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ede76a8b062a23466f47d82d68ab4f3e3595402c63648fecf7f00ed772026126",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:44+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_zuco1.json b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..5aa54171d90b6ac70f121de1ba90c1989f50f8a4
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_zuco1.json
@@ -0,0 +1,130 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1.4b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1.4b-deduped",
+      "revision": null,
+      "sha256": "6d145402b261c9ea294f8015f76b244e40465f445e8d16f7789dc201da8be4e5",
+      "files": [
+        "config.json",
+        "model-00001-of-00002.safetensors",
+        "model-00002-of-00002.safetensors",
+        "model.safetensors.index.json",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e1ab6d6dc71bd6d1e2b65b295ab5b8996a93a04d5b8285bf963d5a9fafb9ce33",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:53+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1.4b-deduped_logit_lens_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "ede76a8b062a23466f47d82d68ab4f3e3595402c63648fecf7f00ed772026126",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_celer.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..d32b191a2f0f387dd61b2076ab5599e04c100bc7
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_celer.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6096c241a3fd9ff51063d7e5e241e0374ed657476779b1eac30d388d3461fd9c",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:23:10+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-14m_logit_lens_surprisal_tokens_celer.pkl",
+    "source_sha256": "8e8fc1180c64bd6a3303715660791132cf643c52a7900e29cb964cfcc5e50c66",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_dundee.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..14dd7f891477c2ad73900e2b9a4038822fd9801a
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_dundee.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "02410790212654464aa97caf162b66c52d45230efa3b241085e10c087aadd4bf",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:23:22+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-14m_logit_lens_surprisal_tokens_dundee.pkl",
+    "source_sha256": "14d887374392a36ada76b003111e82846c9ad2cf45d5439586ef2ec472fa457b",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_geco.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..bfa17a9df1cae778086b2a60c7f6c34de3fbe5e5
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_geco.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "81eb0dbc66b0aa51a6f86089319a2e203bd2119a4e31a4257c3380b99e2885d8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:31:14+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-14m_logit_lens_surprisal_tokens_geco.pkl",
+    "source_sha256": "53980bf06997a049d382d4db60b35c53e3cd39a7dc0d158d75fa7235936b7cde",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_onestop.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..50d760098b629f5dd84e2e3f2846a3c0b3fbf6c9
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_onestop.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5d52ceed2b9c70ef1ca0b7d610202636d4f0a24d0d45d03e8eb3d7f164932a4d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:29:26+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-14m_logit_lens_surprisal_tokens_onestop.pkl",
+    "source_sha256": "8036dc462aedb36f89d59da3164b1c70b27a8eda3759d57120e1a1cee0ee1ea6",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_provo.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..923aefbc2a7e59c7df1cbc2f62030e542d78a282
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_provo.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9cb06eae8f88ed36b4f2b950ce38636c0f49f094221ae9259b9d6e64ebecba7a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:09+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-14m_logit_lens_surprisal_tokens_provo.pkl",
+    "source_sha256": "bbd52152baa626f8967ee44e1f57914c0dc713eee4ad56f858edc603a126bbff",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_sbsat.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..00ff6d25ee7e045cbb76bebfa179105126e11bd5
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_sbsat.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b85cc89381cc6b247eada7016778912d771c7c64f435ea554e94530d5b77e07a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:01+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-14m_logit_lens_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "d07bdb6bdeb96957af3d66f915f2d491c2b3a8c666c17f6dcc421a75d5c5700f",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_story_sbsat.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ae6ec45a9137de56d62cc719915c3c95a38494f
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_story_sbsat.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "954a43bd82e7acd3755c512245ec2165694212c1e449482905bff44411ad54ab",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:21+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-14m_logit_lens_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "272f5c8454432d63623972ec8c4c0f44b5e7fac14bad4a186af8b81e8b3044b5",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_celer.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..04557cf90783e1a64abdb9823c6156fa464e0205
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_celer.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8e8fc1180c64bd6a3303715660791132cf643c52a7900e29cb964cfcc5e50c66",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:22:49+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_dundee.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..a49aa36e2292bd124e5d57af9c3133b5faab84ce
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_dundee.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "14d887374392a36ada76b003111e82846c9ad2cf45d5439586ef2ec472fa457b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:23:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_geco.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..50c234a30a8314b4947eaff392cd7c9dd8240d35
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_geco.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "53980bf06997a049d382d4db60b35c53e3cd39a7dc0d158d75fa7235936b7cde",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:31:01+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_onestop.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..50b1eb8f1293528ffd3f651c9c9ece841ca57f4b
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_onestop.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8036dc462aedb36f89d59da3164b1c70b27a8eda3759d57120e1a1cee0ee1ea6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:29:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_provo.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..a64a0040c1a4504a7ff4e0d87297b99630c5f297
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_provo.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bbd52152baa626f8967ee44e1f57914c0dc713eee4ad56f858edc603a126bbff",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:01+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_sbsat.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..1555fa79700574271ffc0c09ffc6d9328bf027bf
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_sbsat.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d07bdb6bdeb96957af3d66f915f2d491c2b3a8c666c17f6dcc421a75d5c5700f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:00+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_story_sbsat.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..f21952043ceaf214b9f1e804e8920c79244dd418
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "272f5c8454432d63623972ec8c4c0f44b5e7fac14bad4a186af8b81e8b3044b5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_zuco1.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..8aceed7f7c861dda6d86d54f14e88c9f859c5368
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_tokens_zuco1.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f3e240b2f048bc13a0a444c3310a75ff121ce38483eacb2fbc44d7787a0e741d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:08+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_zuco1.json b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..454651e9a02a61161627df67222b33af625bb1ea
--- /dev/null
+++ b/manifests/EleutherAI_pythia-14m_logit_lens_surprisal_zuco1.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-14m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-14m",
+      "revision": null,
+      "sha256": "0e774dd3f89ac7d724dd3c7b41993d277a008e97fcef8c9e5ec613cd16a87215",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9b7c381b069d1beac94e2a934064054d74cfe1b1765180afa257033a4ca8a6a0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:15+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-14m_logit_lens_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "f3e240b2f048bc13a0a444c3310a75ff121ce38483eacb2fbc44d7787a0e741d",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_celer.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..eeda3fc8896d675b9b105c5da3736d6ae2759b20
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_celer.json
@@ -0,0 +1,104 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "10e1c449ab216c87e8970cb6820be3cff2b3be4b1145ac3994364b6a369ab553",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:26:33+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_celer.pkl",
+    "source_sha256": "5053c8395111bc39747aa935173d27ac5c528e7ad4f04ab92b8575b16b13c23b",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_dundee.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c4cfda41661ba839304d0f4b45f16631cd3d3d7
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_dundee.json
@@ -0,0 +1,104 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5c38cc0d2f50a61794860532fe582778ce9701ffc7a32616e62e4628eabf1590",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:23:53+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_dundee.pkl",
+    "source_sha256": "9e46a86ead8d1606e8056a26fa146264e7f860a21c6946fe577e400625f4129c",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_geco.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..26be89585ee48fa3e6c6b3c9e64340e10f612173
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_geco.json
@@ -0,0 +1,104 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c58d3fcbadfbcd2dec111fff402543b89f8ec8f4d6944abed779017618628200",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:34:13+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_geco.pkl",
+    "source_sha256": "7e2f3bc2957bf8aefffb933915b426050dc6c528748ab50e432aecc13694d339",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_onestop.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d4cff0d43c94c2c4d65d145c9d50a9814bf89d2
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_onestop.json
@@ -0,0 +1,104 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "285e037ec928b5f85e7f4134d23a78cde971390189f0fa8d391e1f6c4dace8bb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:31:02+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_onestop.pkl",
+    "source_sha256": "eaffaa5016d5d20041d64f6b5bca0aacf7a7bb231170311c78c3625ad13cfedc",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_provo.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..747cd49d0d992c759afdb256e6b6797b8a662d8f
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_provo.json
@@ -0,0 +1,104 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "52531ea47800a12d68de50eec4a9b55111f118f90d1cde8af9059155ec61241e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:09+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_provo.pkl",
+    "source_sha256": "2121e51cf1b4b51ed07f8c307d6cf57b44f46f982cbb91153b137974d09d743f",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_sbsat.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..0296a1ac3183c1dbb51c670d5575f1df3e77365b
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_sbsat.json
@@ -0,0 +1,104 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "025fcc14336eda08a80853c114486001f9fc54e12fe8961b0675679ae777b34c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:13+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "429d573e003d0256cc634add1b131e04540132ddee5657062abcbc884894ba3b",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_story_sbsat.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..e79713392317d7162e68ac76bde8be268726c261
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_story_sbsat.json
@@ -0,0 +1,104 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2a5f666e386b9ff0ce759c12873820aa8833c9ed63e729d175adaf59628342ef",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:27+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "c0f98228753c9a19e1d03b0f45b75dd295c2971924481d4d7f83f1411b466e20",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_celer.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d4ee4fd11fc1c9423212166e3b39f3db29ff03f
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_celer.json
@@ -0,0 +1,87 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5053c8395111bc39747aa935173d27ac5c528e7ad4f04ab92b8575b16b13c23b",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:26:03+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_dundee.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..32b82779ee866bdf7731416b980b5a40fe43a7ee
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_dundee.json
@@ -0,0 +1,87 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9e46a86ead8d1606e8056a26fa146264e7f860a21c6946fe577e400625f4129c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:23:42+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_geco.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..9906959105f3454906e9cb4c8f3ad7ea23aebdf1
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_geco.json
@@ -0,0 +1,87 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7e2f3bc2957bf8aefffb933915b426050dc6c528748ab50e432aecc13694d339",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:33:58+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_onestop.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..edd3902891b7551d07df8ae88dfe3c135b6afd13
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_onestop.json
@@ -0,0 +1,87 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "eaffaa5016d5d20041d64f6b5bca0aacf7a7bb231170311c78c3625ad13cfedc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:49+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_provo.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a148e7cf0f76053bda4ddb65774c8700dd04d36
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_provo.json
@@ -0,0 +1,87 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2121e51cf1b4b51ed07f8c307d6cf57b44f46f982cbb91153b137974d09d743f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:01+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_sbsat.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..49a2be7609c712378af24134cb671f115b741797
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_sbsat.json
@@ -0,0 +1,87 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "429d573e003d0256cc634add1b131e04540132ddee5657062abcbc884894ba3b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_story_sbsat.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..24c701d17adbcb0d5b44c58304772fd8ec84c49b
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,87 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c0f98228753c9a19e1d03b0f45b75dd295c2971924481d4d7f83f1411b466e20",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:17+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_zuco1.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..40f65bae849deee16bb4f028a183d4f2b86daa73
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_zuco1.json
@@ -0,0 +1,87 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2cc2526ede45eb0c1a1f18037b86fb01c9abb8274c3c1252cc1e5f6cec453491",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:36+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_zuco1.json b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..34c072c5faa0c0d3eeb768ed43fa9163d7a43840
--- /dev/null
+++ b/manifests/EleutherAI_pythia-160m-deduped_logit_lens_surprisal_zuco1.json
@@ -0,0 +1,104 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-160m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-160m-deduped",
+      "revision": null,
+      "sha256": "711c102c5b9738ccdd85101c29ef13def9dc434097a690e46b319bc40e2b0130",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "da3bdb79a0f2ef224960a9d9347311bcdb87ab08f0805b899119003529102c48",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:43+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-160m-deduped_logit_lens_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "2cc2526ede45eb0c1a1f18037b86fb01c9abb8274c3c1252cc1e5f6cec453491",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_celer.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..767cd54565bb616e76c762c2b6ea7e7c7daac956
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_celer.json
@@ -0,0 +1,112 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "45dde0889ce3e6a858cfed08504684e453c7a407d1630cc037e549a752b20657",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:11:13+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_celer.pkl",
+    "source_sha256": "93c6648595299694791ed44a2b64365818cbd24202593508fa0e28cfbb6dd904",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_dundee.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..8768aea6413dceee79e437f553b6852e50cc557c
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_dundee.json
@@ -0,0 +1,112 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f6fb2880dd3762a77b0d67f0ad83c57e80a67f24a2732956647db01d1edb5e60",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:24:21+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_dundee.pkl",
+    "source_sha256": "0e8aef361e0add64679d7b56ecf3f80425005759fd4ed5fa865bb2892368e95c",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_geco.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..8664a4559ed93dc5bdd369d7531f74e20d840471
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_geco.json
@@ -0,0 +1,112 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3117d6031da792c640677a0b370d1502e3d9b3e84d0ef69ad20073f3630e8b09",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:36:21+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_geco.pkl",
+    "source_sha256": "05856dca66e825ff05502138285bf55612c006c6d0a87947aea2cfceb3786104",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_onestop.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe4d1b0013c31dd06b995f628092bff0f6dc5adf
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_onestop.json
@@ -0,0 +1,112 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c6c0f2a212783a5c45346aedc8d25a8f8bab73d45a9d29342916c3daa456480c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:34:01+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_onestop.pkl",
+    "source_sha256": "229e0c448885543a2137636c92bd5445946508a9ed37b02d2eaf55493c21a556",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_provo.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..3562a3cf8c3ea08e9aa985c02bf519929b453050
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_provo.json
@@ -0,0 +1,112 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "46ad0028a33e740f3d5e51b90686dbfab75c8eb8ddc2c02476a97403d4a1c994",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:26+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_provo.pkl",
+    "source_sha256": "93e9b7f45d7708593e785f8d0fb706331f4179cecff23042ad002f4aeaf13201",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_sbsat.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..06a63b926264f2c300e0d4b9decce1ef7903da8f
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_sbsat.json
@@ -0,0 +1,112 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5755d1aebcebf2ba001cdea4fe3550ce96fdcb9d29dd6be25078cc0445729113",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:44+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "9a19139412ed95c78851e4146ade223718b2e58f287a3f3f0e6c59e73c3a70e5",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_story_sbsat.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..fef4a088d2e91ce10c5ab61424eb57be513f4ef1
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_story_sbsat.json
@@ -0,0 +1,112 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e3b67b80fe4354ccbb2887cfe4e040b9140a488e444e9f4eddb0123eb3e08bbe",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:20+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "e8defd10211fc52f6a0dc9c303e862e83b0bd17908d42922c76f04b8754e2b00",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_celer.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..ad543384d7834c61d9df5c5088ae31461bb3df57
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_celer.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "93c6648595299694791ed44a2b64365818cbd24202593508fa0e28cfbb6dd904",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:10:37+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_dundee.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa22de24e1de22d8bae46df6947d095ddc8d5ebb
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_dundee.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0e8aef361e0add64679d7b56ecf3f80425005759fd4ed5fa865bb2892368e95c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:24:09+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_geco.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1e50030e69d5e789aca79baecafa7178b3f129c
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_geco.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "05856dca66e825ff05502138285bf55612c006c6d0a87947aea2cfceb3786104",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:36:07+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_onestop.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..635171564f86031ece3c25c9e9b433a53c57da8e
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_onestop.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "229e0c448885543a2137636c92bd5445946508a9ed37b02d2eaf55493c21a556",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:33:46+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_provo.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..406b2b64dc679eacec4bcb478631aeed2ebe7759
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_provo.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "93e9b7f45d7708593e785f8d0fb706331f4179cecff23042ad002f4aeaf13201",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:19+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_sbsat.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ee1985f3e6aa34b12b28f1381c19d647a07fdb8
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9a19139412ed95c78851e4146ade223718b2e58f287a3f3f0e6c59e73c3a70e5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:43+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_story_sbsat.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..b64b803c82b11efc34c027047df1d2209953dd89
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e8defd10211fc52f6a0dc9c303e862e83b0bd17908d42922c76f04b8754e2b00",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:13+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_zuco1.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce49c4bfb0418281a6b0cc10bdd8b67ec3148bb6
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_zuco1.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9b2a0ff0ee2c82bb36dc048ae8a21f7da7776cb4fd305f32c1aaebb64434437f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:46+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_zuco1.json b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..54cc1cae703c70712a2297f9e0f15c00b2466676
--- /dev/null
+++ b/manifests/EleutherAI_pythia-1b-deduped_logit_lens_surprisal_zuco1.json
@@ -0,0 +1,112 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-1b-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-1b-deduped",
+      "revision": null,
+      "sha256": "aa1006d530f95137769575013afc5d773630c8be6a5836b982d512277337c03e",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b80ddc1c03c720e21b61e37ed97e9a0acc60d7ba5bf99d82b69203b02d875dd6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:56+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-1b-deduped_logit_lens_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "9b2a0ff0ee2c82bb36dc048ae8a21f7da7776cb4fd305f32c1aaebb64434437f",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_celer.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..c07e0949fcda3babb9aaf94cfacd72ac92ba13f6
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_celer.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a9f824d612a1b0eaf05274c3fbb4481144a6cb3c4867df4d62324317eec948de",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:24:49+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-31m_logit_lens_surprisal_tokens_celer.pkl",
+    "source_sha256": "633032379a038f06cd9def2b17aa1e5229ad1ffb2217f5ce74a73ae5afe99f1e",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_dundee.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..b8275d7045deddb50c67af76e00ff938d7111322
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_dundee.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "dc61e91c38aa49fbc4cf2a96a104a2c0e6b5c12d479e079b2d7c2da7687eca51",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:23:10+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-31m_logit_lens_surprisal_tokens_dundee.pkl",
+    "source_sha256": "ab16e6f9ff39cdcba2ab5d368cf9096659820aa4d1ce0d2d4a219ef187d1f019",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_geco.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..c88c6284b39904570a6b477939d305b025750a33
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_geco.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9f12cfc27b7764d1d74082a036097202954ab47a5e8df8bed6e10dab2de2dd0a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:31:15+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-31m_logit_lens_surprisal_tokens_geco.pkl",
+    "source_sha256": "f6418f3d42f1b0764170c42ab20da18feada0885bb3b91dc16645f8ed9648521",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_onestop.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..db00cd385a2a507670c7938da7cb4997acfb4944
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_onestop.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8a5ec88c3fc0e29e98c0239a994a8441568c76dea02adcf63264ec3b729f60ce",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:28:31+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-31m_logit_lens_surprisal_tokens_onestop.pkl",
+    "source_sha256": "234a79cde1b30e4428da1ce61cdc6c474819ac78bf53d0d9230e5fa32f421796",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_provo.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..9cb982c692f690a3de88065ba4e9aa9a53f356c3
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_provo.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "dc150bb5ba61645aecd01ab2cc5b08c830e544ab7fabb035e1c89a0f4b53f179",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:24+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-31m_logit_lens_surprisal_tokens_provo.pkl",
+    "source_sha256": "9307b4c8aa3c73f8995cb3d56c13d5affa52ee4c2120fc96e375f43853a93bd9",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_sbsat.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..d4598f93bb15a21034c6351704efb4616ecf1235
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_sbsat.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b84d13d0647114a015522e179dad9c2abb7b68d440103bb4a093d047579a1e91",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:37+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-31m_logit_lens_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "81cb028afad3f8f1c78931bf0fa3503436c45b1b573ca24b5653c7a6dabb2d25",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_story_sbsat.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..0036b5ad3659ba7645487c67404d2fd9cb51477b
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_story_sbsat.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fbaa54cd678488ce5e2d65db4a13cfe5211868c275764853768b6d201516b1af",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:03+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-31m_logit_lens_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "18e720a909de5bcd3885064b4c7d8a3169981721d74025c464279ff69ea7d944",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_celer.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8ea055b9929ce63dc254b167076b497d54f7088
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_celer.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "633032379a038f06cd9def2b17aa1e5229ad1ffb2217f5ce74a73ae5afe99f1e",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:24:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_dundee.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eb14f327e2348cd7875876025d36377bca2363b
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_dundee.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ab16e6f9ff39cdcba2ab5d368cf9096659820aa4d1ce0d2d4a219ef187d1f019",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:23:00+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_geco.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..79f850f8a3a4334268ad9a003b080c8a8a37f0bb
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_geco.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f6418f3d42f1b0764170c42ab20da18feada0885bb3b91dc16645f8ed9648521",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:31:03+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_onestop.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b207eed12851c264988a6b5576f08b02899da10
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_onestop.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "234a79cde1b30e4428da1ce61cdc6c474819ac78bf53d0d9230e5fa32f421796",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:28:22+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_provo.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..dce1cd2f17906f07c3365af650ae6986ccfe56e6
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_provo.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9307b4c8aa3c73f8995cb3d56c13d5affa52ee4c2120fc96e375f43853a93bd9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:17+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_sbsat.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..707f88cf18a1e71ea810cd017b788f511cf69b73
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_sbsat.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "81cb028afad3f8f1c78931bf0fa3503436c45b1b573ca24b5653c7a6dabb2d25",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:36+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_story_sbsat.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..6404105f7d74d7f01d5943c45e78bc09233faed6
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "18e720a909de5bcd3885064b4c7d8a3169981721d74025c464279ff69ea7d944",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:26:57+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_zuco1.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..d1115a74b7fdd656ec6553ae17b58df106b60c6d
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_tokens_zuco1.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7f20d15632e1fdabae9a1e568abbfe76b55cb8bc599e71ffafb74a80c02b2d08",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:27+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_zuco1.json b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..48a84f9e094b5499bc499154f4a0a5dffa237bea
--- /dev/null
+++ b/manifests/EleutherAI_pythia-31m_logit_lens_surprisal_zuco1.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-31m",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-31m",
+      "revision": null,
+      "sha256": "603b121ba43fde5038971c9063600636e25adf41e87c497ca649fad8289e4ed9",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2d1294db05fd98656a3b018fb0b3f33986cf317f3d46991b35bdde7c35134517",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:34+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-31m_logit_lens_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "7f20d15632e1fdabae9a1e568abbfe76b55cb8bc599e71ffafb74a80c02b2d08",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_celer.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..f7c953c54c8b34950dd1e96c7d83a4d4ecb41beb
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_celer.json
@@ -0,0 +1,128 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c3c46ac3a424d8de12d64718503713fad781867a5625021928ef88e82ee7cbad",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:25:10+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_celer.pkl",
+    "source_sha256": "439516afc7dd0b3f2bf343c8f079b4dc4beff910df0d39861275e2934e42837f",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_dundee.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..22e6287fd6aa56815fb28aabf740b9f17eee4a6a
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_dundee.json
@@ -0,0 +1,128 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7854b07e3635d9065cced1ffa99d5530a2ffbd06200bac2c109f4978bc1faf46",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:24:50+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_dundee.pkl",
+    "source_sha256": "7a4a2aa08fc7ae471e21f93f5d655d3f616373097858d265085c193f695c91c1",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_geco.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd6cb0a5235797936e03f0c525d0fcc2f87c186c
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_geco.json
@@ -0,0 +1,128 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0da34d1509fa15b321ee96f4921340befde0f04212b574df515a301aad317907",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:40:08+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_geco.pkl",
+    "source_sha256": "87691484b87360e57fa007fc8213660c02406ce550793a4b4f7f209497597616",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_onestop.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..d80c78c2f27b5c8d126fb58788399c3d20101747
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_onestop.json
@@ -0,0 +1,128 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7b11a4fe736a7b91aec91c6e5f7411f5a2a425e5591bf8644247a36052f9f04e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:33:30+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_onestop.pkl",
+    "source_sha256": "478ecb5ea58dd86c623e1ad69042eb3cfe81752406d5b393e921889d82019f4b",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_provo.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..d83c4d6c13024fdf7dd19354803fdd631ac0a361
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_provo.json
@@ -0,0 +1,128 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c7806d07248d368d7e04c91eae70c89a536253a22c7ce5ebef6c695deb5d5e35",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:30+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_provo.pkl",
+    "source_sha256": "0996cee27859b8d185010c45214d41cb65762bd64e48cc512f37593be4a7d4e2",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_sbsat.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd009c73f9f2b96aabb99e4351e0f905942e32e2
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_sbsat.json
@@ -0,0 +1,128 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3c2f6e066b65bfbfbb1186af7af199ecd2e40d1328b7f940c954450658265ead",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:59+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "7bad7932c15fb70db80fa3b7807bbc840e736daf2ef323f3813c2bcafe325290",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_story_sbsat.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1597817095aee66251f06d52df8b904c9edef4c
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_story_sbsat.json
@@ -0,0 +1,128 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3d1feb020b067148e0cbf99fa5632c722c82a0e40cf3efbd790f396693ae28dc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:12+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "9b5f053c08a6064e1b22310f10e2acb9740940de7d464b87a8d2c912d292d501",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_celer.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..5961bbcf77d971f0330e88e80d5b924ec73f949f
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_celer.json
@@ -0,0 +1,99 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "439516afc7dd0b3f2bf343c8f079b4dc4beff910df0d39861275e2934e42837f",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:24:20+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_dundee.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..7cb74460956e0469cf65f0287215e6ffc3c2e39d
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_dundee.json
@@ -0,0 +1,99 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7a4a2aa08fc7ae471e21f93f5d655d3f616373097858d265085c193f695c91c1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:24:33+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_geco.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e2b2fd6fd8d79b45bfe2facf98b02f428f7318e
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_geco.json
@@ -0,0 +1,99 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "87691484b87360e57fa007fc8213660c02406ce550793a4b4f7f209497597616",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:39:49+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_onestop.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef48cb720966266ce0a3b25f9c9066c257cb1dd4
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_onestop.json
@@ -0,0 +1,99 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "478ecb5ea58dd86c623e1ad69042eb3cfe81752406d5b393e921889d82019f4b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:33:17+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_provo.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..08d083a8fb6685df172e78048e530702a87e03f4
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_provo.json
@@ -0,0 +1,99 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0996cee27859b8d185010c45214d41cb65762bd64e48cc512f37593be4a7d4e2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:23+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_sbsat.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f984fa2a0f7d4688a0e6fe96939685ce89f515c
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_sbsat.json
@@ -0,0 +1,99 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7bad7932c15fb70db80fa3b7807bbc840e736daf2ef323f3813c2bcafe325290",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:59+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_story_sbsat.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ebf3184418d2ca9279521eb85f1c202edd31190
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,99 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9b5f053c08a6064e1b22310f10e2acb9740940de7d464b87a8d2c912d292d501",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:06+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_zuco1.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca70ab30d4879e7224a4cf6940c667ace129cc56
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_zuco1.json
@@ -0,0 +1,99 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "db898461d615264661dba83b6433a68e3b2ebf1e733a91aa8c2feb42f88dd0a5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:31:05+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_zuco1.json b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4ec99cb4a05a230b44ae7fb271a5c190268d2a2
--- /dev/null
+++ b/manifests/EleutherAI_pythia-410m-deduped_logit_lens_surprisal_zuco1.json
@@ -0,0 +1,128 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-410m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-410m-deduped",
+      "revision": null,
+      "sha256": "388759771a4cd3c5435a61219ee90c0a25a5a0400d93318e208dd135bd3d5b96",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1746dc550ba19914fa1de2a745f4edc63d66c1cbf24b12d4079b2072c14a409a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:31:12+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-410m-deduped_logit_lens_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "db898461d615264661dba83b6433a68e3b2ebf1e733a91aa8c2feb42f88dd0a5",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_celer.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..16cef4870ce58bb4efa0d8c7a54975dea2f53d65
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_celer.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fd2b7f02f04d337a6c701150431c582a514755387a157c2078c49c2fcd632055",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:26:50+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_celer.pkl",
+    "source_sha256": "0d44e0a7c8d69e3cacadc0f47442206d32dad67d98b224fe58f3130c85ceeb3d",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_dundee.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..c14e8ebe79603dd90ec7db5ab52a20af84ecad8a
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_dundee.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ec383fe7c610623cbf4ea01f9375df7a689cef932895fda590957b6fe994cc7f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:23:11+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_dundee.pkl",
+    "source_sha256": "49181834afd42d3134f29d44088043ce11533ad27488f32dece7b7bbf9e6993d",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_geco.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a402e208a2b6a604ff9dc32c12882a7acc163d1
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_geco.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "05a77d4e4798da85a5d34ae63e77aa7b48f67ea56aedd366f36780c24db80e7d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:31:11+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_geco.pkl",
+    "source_sha256": "4dc455a022f9045d8ec2497f26ea2e2032f8d22c664eb4a34e8d11fb822724a1",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_onestop.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a04ba856f6a4dc55835559c63c307a7e32c9eab
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_onestop.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "69adc28dc1e498ce5f7de48f0dd93e4b4b25e5263ed306ebcee0634fd8c01a95",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:29:23+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_onestop.pkl",
+    "source_sha256": "5ac7f51375985c3e9da69d66c50ef23b81a5d09608e1b4fe8afe3260b5b7570d",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_provo.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..69906141f7edb316b9d034a8549d9d3dd3751c1d
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_provo.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a0065f5fd2212ed0605b90d0a03f807328ac9286482b6e26ecc33f0c7f08b9f2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:59+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_provo.pkl",
+    "source_sha256": "291dfbde9e6fd90ccc22d58f8bcb50c84dd7970581a35b5d0de9a5df13cfea76",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_sbsat.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..85388a87ae76205692ac92edeb46bb70c102ca1e
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_sbsat.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d7312cbef7ba6623cf5b52b1fc639761cfcfd7b55005cb0cf3142ef60dbb27e0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:58+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "faf133f5530ef5ba3efe40293193faca6d2d20755bc4aeee8187b7fffd9a8293",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_story_sbsat.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ce226a78162cbcd41f4d7cb28f1c4d1ebaf5cfc
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_story_sbsat.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e48148494d13d4149631e31e43fa78068495762294a8c31bf7d61f0baf7613ed",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:13+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "93b6153a050f94ed987c40f49d954000499234d28c24e7851d40bd0035562060",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_celer.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..6362d248b324dee3106368a725e1f7ceed227870
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_celer.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0d44e0a7c8d69e3cacadc0f47442206d32dad67d98b224fe58f3130c85ceeb3d",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:26:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_dundee.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..5c08221490518040c62049f007fcf3e1dca8bd92
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_dundee.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "49181834afd42d3134f29d44088043ce11533ad27488f32dece7b7bbf9e6993d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:23:01+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_geco.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..5277f7c77b346e15af732ca6bb14e3878f60da54
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_geco.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4dc455a022f9045d8ec2497f26ea2e2032f8d22c664eb4a34e8d11fb822724a1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:59+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_onestop.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..09180aee32026c383d078dc131d0ee00a290f7cf
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_onestop.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5ac7f51375985c3e9da69d66c50ef23b81a5d09608e1b4fe8afe3260b5b7570d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:29:14+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_provo.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..9391ada31dcb172c094c63def8c1e9ca75b2295d
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_provo.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "291dfbde9e6fd90ccc22d58f8bcb50c84dd7970581a35b5d0de9a5df13cfea76",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:52+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_sbsat.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..f6d6d30523de67c8937308f5911c36038203784d
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_sbsat.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "faf133f5530ef5ba3efe40293193faca6d2d20755bc4aeee8187b7fffd9a8293",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:57+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_story_sbsat.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..20490b9ebd82e2ec0756b6867dc1447c9c51e294
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "93b6153a050f94ed987c40f49d954000499234d28c24e7851d40bd0035562060",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:06+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_zuco1.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..dec14a37bc487719d050f57aad2fec4c4418c18b
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_zuco1.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c7c3c5cb614aa7b0671ca1942fe2d3b788aa772a713fffde6457b3f6b5ba94ae",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:54+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 2048,
+    "stride": 1024,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_zuco1.json b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2a58ee09a6b4d1065090403d961f0cafec5217c
--- /dev/null
+++ b/manifests/EleutherAI_pythia-70m-deduped_logit_lens_surprisal_zuco1.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "EleutherAI/pythia-70m-deduped",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "EleutherAI/pythia-70m-deduped",
+      "revision": null,
+      "sha256": "7033c9c2c33fc7a1127bac6d382e2cd758ace20d7cf38fc89a469ac1a2579c80",
+      "files": [
+        "config.json",
+        "model.safetensors",
+        "special_tokens_map.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a0056c4437d9b5c6f45217a798f9d1233a7d4fe54ee5c9a6b693a05eb2184238",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:31:01+00:00"
+  },
+  "derivation": {
+    "source": "EleutherAI_pythia-70m-deduped_logit_lens_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "c7c3c5cb614aa7b0671ca1942fe2d3b788aa772a713fffde6457b3f6b5ba94ae",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_chars_meco_l1_w2_ch_s.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_chars_meco_l1_w2_ch_s.json
new file mode 100644
index 0000000000000000000000000000000000000000..c989c6f97ac61dca1d6a84010a43fff7db5a1c80
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_chars_meco_l1_w2_ch_s.json
@@ -0,0 +1,65 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "characters",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "16cb35d873c0fd1a9516d75a8753aaf2a1bc20a4e53f61888928f151311d0960",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "23c791ca3ab5a4d8a421d8bcc7cd1366cdbb59708edad851b07e5877886ef7fb",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T19:33:31+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_chars_meco_l1_w2_ch_t.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_chars_meco_l1_w2_ch_t.json
new file mode 100644
index 0000000000000000000000000000000000000000..622b9b906b9aa281cc778d459db823cba336fc34
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_chars_meco_l1_w2_ch_t.json
@@ -0,0 +1,65 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "characters",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "5d94c0fdfeaa300075e5381c663ae350ab819851c13e072ec53558a1501ab4b0",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "48eeee21ec7b9aaff0daf7ea755cafbe9e93b576f6daba6f7517f640081884c1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T19:32:11+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_chars_provo.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_chars_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..b6850b95d55df679bbad4e9ab0a58bc09b93598a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_chars_provo.json
@@ -0,0 +1,67 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "characters",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "1bfa5187af0e0c6cc0ae7f014b5ab37e291cbfdb6026b082192345e9680692ab",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bf5eb9d3e6e718c95717519c11d0eec97e28a4abd1759485108d0486bd3ade59",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:51:26+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_de.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..8cbf88f0e429f0bb9308bcad81fedad94737ce6e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_de.json
@@ -0,0 +1,67 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cf174069066664f50567abe9ecb6a0f005e7696ba47fa107769aa53869fae2f5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:38:45+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_fi.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4aced4cac3f4881b2f7e199b8ece851358a179c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_fi.json
@@ -0,0 +1,67 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e96026953653c1b1ee8ed2c7bbe21eb5d92732a16bf0bddbb27b9265c6a7a74a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:50:32+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_gr.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..bfbf62fbfc9191f63c8e5ec8323d5408bcc8b072
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_gr.json
@@ -0,0 +1,67 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0f2ef3d01f0ecf92d7e64b767533419c2df6dc5b37d3c65df7e3cb5ea87d7fd3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:52:49+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_it.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..819943185e61a160c083bfc72a3ad790fb3186a2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_it.json
@@ -0,0 +1,67 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c77b97e60a0a50a3b1884c998b49717a512c8b7ced2564712e6586b51d720bfe",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:42:00+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_ko.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb28d9bb1de2909ba4b000f68c06fccb407ebede
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_ko.json
@@ -0,0 +1,67 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f98b44d2b25712dc03d87831dab2eb454ddb2ef99438bb16ae31fe8a3953b333",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:33:58+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_l1_w2_ch_s.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_l1_w2_ch_s.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2dc4be9449e35e3fbb9f421564ba1b7d0074a23
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_l1_w2_ch_s.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9f27685ea775f0843723c5aa8c5b226b7d3420ccb213afd1eb8198439dcdb6a3",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "401396312a6d8694edbc1f737b2d0da8b6e8c384cb43420901a698b5373b22c8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T19:33:32+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_charbeam_surprisal_K5_chars_meco_l1_w2_ch_s.pkl",
+    "source_sha256": "23c791ca3ab5a4d8a421d8bcc7cd1366cdbb59708edad851b07e5877886ef7fb",
+    "unit": "characters",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_l1_w2_ch_t.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_l1_w2_ch_t.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3484c39c367ba89f53aee4f1a72189d7419b8dd
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_l1_w2_ch_t.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "866884887c127db9b93a7cf7e4deb154d4c2cba960aa58ede0d05493044e3877",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ddd6b9650a08ec259d43f3265806913b8a85c2059086c1b773bde54177e238db",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T19:32:11+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_charbeam_surprisal_K5_chars_meco_l1_w2_ch_t.pkl",
+    "source_sha256": "48eeee21ec7b9aaff0daf7ea755cafbe9e93b576f6daba6f7517f640081884c1",
+    "unit": "characters",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_tr.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f721c8f1c40989cb1472cb62a4c9026df582bb6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_meco_tr.json
@@ -0,0 +1,67 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a38e581c9ddbf964cb973ee87859b7813b4f18026fd805d661cf52c9ddb0ef9b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:52:32+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_potec.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b26e65239c7f84951a024dd29b54bbedd947fb3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_potec.json
@@ -0,0 +1,67 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f668f0de66d04fd53e6b9d3fbd2777d6573261d965ad621d5cd9dcaa4304a174",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:53:02+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_provo.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..a4e1bdfc1139a70653046e3c748c8dbdc88444aa
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_provo.json
@@ -0,0 +1,67 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ede357727edc807455ca0eb457ed82e7443d05c8886c8552e39193c6e0d42954",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:48:55+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_sbsat.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..b09c34993873adc370629a9fb7ad40d532e15bb7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_sbsat.json
@@ -0,0 +1,67 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "51cf80deacc11ad7bcdb2554842bf74bacfa4b68383ddab49f0472d88e30d5e8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:34:51+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_charbeam_surprisal_K5_story_sbsat.json b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..fa6725018ffed30975f602df91932f52505a12fb
--- /dev/null
+++ b/manifests/ai-forever_mGPT_charbeam_surprisal_K5_story_sbsat.json
@@ -0,0 +1,67 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "72607462220b4256b5cd53a5fe34addf3a313441e8f74997493f0e1977d67ce3",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T02:41:51+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_copco.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..4bfcdeb21483699f13393e96ee01cdfeca065b97
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_copco.json
@@ -0,0 +1,72 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "5b7b94115956c171c54fee86749fd04e2002938eef5d34a1f1d0d2b8062c00e8",
+    "n_stimuli": 452,
+    "shards": 16
+  },
+  "status": "completed",
+  "payload_sha256": "8e059b7a2c3f9288a737e21c8452b3ec67bf1ac362a5e5f48e5bd95c8460c2e1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S, Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T06:51:17+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_geco.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..3fbf8f884d0e0d50b0b98174516a3a447c7b5789
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_geco.json
@@ -0,0 +1,72 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "59d726081f2386d0c3033ebea00ed706bd479c1d0269333725308684eea841fb",
+    "n_stimuli": 588,
+    "shards": 24
+  },
+  "status": "completed",
+  "payload_sha256": "884fde39d121f61cfb8660f9df495b1e1d4ac65d4b4fbe41ef275362c6eb1e07",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S, Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T05:53:32+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_de.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..07ecedf99f1380c131c46d5914ada34d7f74ce5e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_de.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "77bbe0e30e08c5b7b93e09ddc623d28f309ed1970bd987bb160168e7a4bae308",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:57:57+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_de.pkl",
+    "source_sha256": "945d2218c93fed312f8adb8bd05e6d47f1327496c58ad3fa78ea4177af7cb1b4",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ee.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..c20eceaf7aad794913f767c63a698da4712a7ae4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ee.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4f37e9ba39258d84a84201c26f040b2d9573a83cef8b37ebdfcc7d37178b8a0a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T20:25:59+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ee.pkl",
+    "source_sha256": "1b18f015dbdba78a0438f493fa30a0b4afbff37197bf172f73ce834d436d0679",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_fi.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..d569c91099ec4a0925dc93f0f67ca5c07efee5d0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_fi.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5795c3896dacfc514c04f7c318252d8d50158e924ce6425925762c37352f5404",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:32:08+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_fi.pkl",
+    "source_sha256": "c783c75ff9342a4011a598931e570d654643e6d1e1138da2a388c4ee37aefefc",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_gr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..aac221918d957aff15001772df4d6b1deccbc135
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_gr.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4f7cf9b548fd396a25796602499c5e4e9149e9336e2ca0ee68073369a71acf13",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T21:03:21+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_gr.pkl",
+    "source_sha256": "3b9bec1bf37fabca7581c55f8a803145972e9d89cddb22438a9b084910f331a1",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_it.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..d614c3b767fbac6ab978fa8d773c99852c736b07
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_it.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b2df59e6b701c25d9398638b8d98fd5c88b65ed49f68637cf0acadb7256c6cf3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:10:24+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_it.pkl",
+    "source_sha256": "25fac561fc14aee8e54507fac8e8da55e5efd2ff137e21fa75b07a4420ada3bb",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ko.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..b554c01a710b59580372e3450f13677df89eb1a3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_ko.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "710a91e7e03fccb6f2a9acf02761533ca62ec0d5935b400cbbeaca961ffa3dbc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:59:25+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ko.pkl",
+    "source_sha256": "3d4bf569ff869f4a134268e0728ab9131f20e9f541ffbf223a259b5a81f51416",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_tr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..aad150de5b21dcaedd50bbf72f4fb206e6ed484f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_meco_tr.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "013be368bc8739d8e8bd4dc7d6885adebe3ce1711288f3d3ed87f47bb9484ac9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:08:13+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_tr.pkl",
+    "source_sha256": "fe300b2f9c886b3a5aa1e1bf09268fc231c950dfd4d004b145d09873514f2b89",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_onestop.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..14266854b891af58e7f69203a4361c76fbc734f3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_onestop.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7a8d5aea363be2043ae40ab1fcbaab4e4ad5238a77210d409e677bf269ea297e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T11:46:00+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_onestop.pkl",
+    "source_sha256": "77be07f1ba2d84c1a5f94deecb529c0b1aa7f48be616a363f4278a62b032b047",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_provo.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..25f77b1bcafb4e4c1a7855ecef0b45fc6859ed8a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_provo.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fed81250085f91d82931b13d821d465172c946ddcb584a3a07274e7deb27f1ce",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:46:20+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_provo.pkl",
+    "source_sha256": "97274c10f7d23a4d5d6d10c311fbdaa35f3f2bc86dde9324a2543177d8aea7d3",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..e9e70e92e33bc27cc55ebaacc67eea28458fefe6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b13c0b4cb266ad95b3259d0a1d58915fbff66eb7f49d7373252e98e3b6d68f3a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:53:16+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_sbsat.pkl",
+    "source_sha256": "72d5d52e5d695b2248058d0ce9ac71338294075f03396c88e9d5dd13f0f9e647",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_story_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..f7183f38107d770512675678b3391c9b448b1008
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_story_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "84c5ce212164a61d6ce85623a5fd5a709dd6a4bf9e6ce2c9914dee0a64105e95",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T12:38:20+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_story_sbsat.pkl",
+    "source_sha256": "fb821ea41fd2155e98665260f52307c788c00002a7e414ae41f8ac4cf9925ca7",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_copco.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..90cb720c2d03071ea6e1e4a82e0f37e36a71b998
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_copco.json
@@ -0,0 +1,72 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "5b7b94115956c171c54fee86749fd04e2002938eef5d34a1f1d0d2b8062c00e8",
+    "n_stimuli": 452,
+    "shards": 16
+  },
+  "status": "completed",
+  "payload_sha256": "b1198ce314872879723f7f5ac3f77391f4040949c75d65d05e0fbf9166664f06",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S, Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T06:51:15+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_geco.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..efda19e956351c0ba848754a9610a8692d9d29c3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_geco.json
@@ -0,0 +1,72 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "59d726081f2386d0c3033ebea00ed706bd479c1d0269333725308684eea841fb",
+    "n_stimuli": 588,
+    "shards": 24
+  },
+  "status": "completed",
+  "payload_sha256": "b22b0a2e2fa017561e806b5d65db904dc19f568fb2e3598789665df657555b1d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S, Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T05:53:30+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_de.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..22b5add2500dec24bb42a30b81e778ffe62fe593
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_de.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "945d2218c93fed312f8adb8bd05e6d47f1327496c58ad3fa78ea4177af7cb1b4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:57:56+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ee.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..d0a4bbd5fee28d36cbc712126b3024b6e8e0045d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ee.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1b18f015dbdba78a0438f493fa30a0b4afbff37197bf172f73ce834d436d0679",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T20:25:57+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_fi.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..45aebd50363bc3c49f27e8efbcf04135f624c2a8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_fi.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c783c75ff9342a4011a598931e570d654643e6d1e1138da2a388c4ee37aefefc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:32:06+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_gr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b3c24f2fbd5bba9b02588d90c6ea6bd809af6fe
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_gr.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3b9bec1bf37fabca7581c55f8a803145972e9d89cddb22438a9b084910f331a1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T21:03:20+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_it.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..4cdb696f20460be8021b7b50ef5cb5c0809204c4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_it.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "25fac561fc14aee8e54507fac8e8da55e5efd2ff137e21fa75b07a4420ada3bb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:10:23+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ko.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..686963c2f2cabf742923e54c045712870e5ef387
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_ko.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3d4bf569ff869f4a134268e0728ab9131f20e9f541ffbf223a259b5a81f51416",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:59:24+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_tr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..14b9b82dcb2c7bd11087344d33a5f64ac8efac25
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_meco_tr.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fe300b2f9c886b3a5aa1e1bf09268fc231c950dfd4d004b145d09873514f2b89",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:08:11+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_onestop.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..23edb1d4b3681bfd8fa4f0a4c0859c607451e7e5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_onestop.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "77be07f1ba2d84c1a5f94deecb529c0b1aa7f48be616a363f4278a62b032b047",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T11:45:54+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_provo.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..d76cdce8eb77640bbf43a9845462f845c5b625fd
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_provo.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "97274c10f7d23a4d5d6d10c311fbdaa35f3f2bc86dde9324a2543177d8aea7d3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:46:19+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..06eb7fea514bd9720cdb7179b59b4046c3fba488
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_sbsat.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "72d5d52e5d695b2248058d0ce9ac71338294075f03396c88e9d5dd13f0f9e647",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:53:15+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_story_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..74241cf3db693d06330f56915d27e69e96f71cba
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L10_s0_tokens_story_sbsat.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fb821ea41fd2155e98665260f52307c788c00002a7e414ae41f8ac4cf9925ca7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T12:38:18+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_de.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c218fec2e250d957d0338431bf2e6eefece7f4f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_de.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4097278bec5ac8525ab7fe2c480df5a2dca0a074dcebde4f535b108f17545afc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:59:41+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_de.pkl",
+    "source_sha256": "e9a95ba26f448afb269788c693cdcb81c83d3be0a8cfb163e11f701f5148afae",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_fi.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..5957e04b4daa1217c65af092d5cc37edc6c2e936
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_fi.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": 16,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "86de3da6651f7dd23a4541beb1de2495e66ec4e1176c8a09b08c9a781ad25c81",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "422566ec85121d80f066d8c9d4e1b4a61493d55a7bdfc912850f620dbd6089ab",
+  "environment": {
+    "code_commit": "83e7c3d7f61da81e3437396fbc61894823d3cc46",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T20:49:50+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_fi.pkl",
+    "source_sha256": "c5ab319c9c5b760145538ee74af63a01c855d2c0a4147323c9eed77e2aed3366",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "83e7c3d7f61da81e3437396fbc61894823d3cc46"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_gr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e8754fc57a19aa1c3f797b06d17f29bbb956947
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_gr.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f0e4a88492b027b2e48bc31de4516e17e8236996526a380e4e68d2a4c164ee20",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T00:39:00+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_gr.pkl",
+    "source_sha256": "21e6bb1b31e1fd1700ced4da75569ece1d4260bff959ad5685a1d63ec1ab9ec6",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_it.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..b686e851a44d10cafcec0b0de93f6c5aae8e65fe
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_it.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7406eb660f4c615d013a5355aefd31ce6d0094c6fea5cc3add858eb5f82935f7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:06:00+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_it.pkl",
+    "source_sha256": "1a5269878563e6adb035d4b3df9ae682a78c3bdffb3a358fcf66d1ffa41440f2",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_ko.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..47ee00428d6de68b5534833b27c92f71298bfa57
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_ko.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "397af9ed0c3d6ec865f78da80c48f1a252315e7b91f73eeaa8a145d748a8adac",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T23:08:33+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_ko.pkl",
+    "source_sha256": "f758fa99ea2fc259b207959b371046b8d7938a73b0132372e5ef0db985d809e2",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_tr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..91fe1faf2ec4775c9a42223589e70f3be2cb6f51
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_meco_tr.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6db36c3f3c9971b142d33d71b90312367023092b440ea3a58b660236b3ab902a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T20:56:38+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_tr.pkl",
+    "source_sha256": "76e52aa1208cfaaf3c9da550bcd68cf2ac30bb180dd9f8eb87c4cf8ad52836d7",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_de.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..9ca819f044983c6b10d48cc3a0941ce74076b4cf
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_de.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e9a95ba26f448afb269788c693cdcb81c83d3be0a8cfb163e11f701f5148afae",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:59:39+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 241,
+    "continuation": 15
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_fi.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..330e5f1d6decc471c9e5c4e5a323022b85646878
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_fi.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": 16,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "f21f9cc20f8b7864a4e4b6ea979ed47324ff75009dd11aaaebe027bfe8f42c03",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c5ab319c9c5b760145538ee74af63a01c855d2c0a4147323c9eed77e2aed3366",
+  "environment": {
+    "code_commit": "83e7c3d7f61da81e3437396fbc61894823d3cc46",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T20:49:50+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 241,
+    "continuation": 15
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 16
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_gr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4eafb551525abf979a6cf7aa77f78f1602daedf
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_gr.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "21e6bb1b31e1fd1700ced4da75569ece1d4260bff959ad5685a1d63ec1ab9ec6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T00:38:58+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 241,
+    "continuation": 15
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_it.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..327a9cb9662b7b5f6816aa437826476565be7b22
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_it.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1a5269878563e6adb035d4b3df9ae682a78c3bdffb3a358fcf66d1ffa41440f2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:05:59+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 241,
+    "continuation": 15
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_ko.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae0f5e94716ca6b8471d85defdd3ff86f151a579
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_ko.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f758fa99ea2fc259b207959b371046b8d7938a73b0132372e5ef0db985d809e2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T23:08:31+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 241,
+    "continuation": 15
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_tr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..e44c6ef5dc0ade353e3e457520c781527d09e4e2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L15_s0_tokens_meco_tr.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 15,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "76e52aa1208cfaaf3c9da550bcd68cf2ac30bb180dd9f8eb87c4cf8ad52836d7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T20:56:37+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 241,
+    "continuation": 15
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_celer.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..8fef81cd50f986b734cf20641326c3b8b51a1452
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_celer.json
@@ -0,0 +1,72 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "9a8c9a95b7231e36b6cfb33e3ed9e86c022f56200a900603a3f42f5f27a01b12",
+    "n_stimuli": 28535,
+    "shards": 16
+  },
+  "status": "completed",
+  "payload_sha256": "135bb86040c6220b453bb7f91ca0c15ad9d5555dd1b4c703f2bdebe1b858a60f",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX, Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T13:36:08+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_copco.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a8a3871f8db76b9055f5db2c2fa4f005d2736e3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_copco.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0b0b080aaf8b9de39815a50d821886fe8f1a8e51c208c63787439fcc9e72c425",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T16:25:39+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_copco.pkl",
+    "source_sha256": "d802b1d294523c261e1eaaaf32a63e9bbd687ff9ae6089c0d430b3c182d35924",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_dundee.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..fde4064dffb615d14314d5c21e2dfaabdfe82bd2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_dundee.json
@@ -0,0 +1,72 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "55e4d5b6adb8f258a2d2e10aa3b9c8ba3cc43942bba3df2f1fe2ec337793c263",
+    "n_stimuli": 20,
+    "shards": 20
+  },
+  "status": "completed",
+  "payload_sha256": "ad456063acecf6bc808b8e77cd9fd4fcb96b68530d09e95b73b188989ed94ede",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S, Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T12:40:15+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_federmeier2007_n400.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ed6ff4d0e1424919a23c40a42f4b83d2c24a0cd
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_federmeier2007_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "456b8cd1658103f4de71af7fd6208a207a0d9f017a3d14e6a33c5def718bb4d8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:34:37+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.pkl",
+    "source_sha256": "43ef1e386fb8e346bc7c930d93e2b4d21c38b1275dde8ad02a9596c97549c45e",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_geco.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5a3564f2f2bd07a0df5e7d8fa8066b469ffb550
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_geco.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cf59a3cc4860c7bb8e2dc4a88f40a94db183d346351ef7e2b509a075b09f37a2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T04:30:22+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_geco.pkl",
+    "source_sha256": "0a1da831d7db529123b3f2c5cc1ea12b1f1b4a3a49e50b8c57b1603d90880fe8",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_de.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..0124dc5f91a441805669918014f108cdb359982b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_de.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "79f15a7900e188528f22c9243e23dfc4c08af14a5223d9dbc4fe5ca2513a2782",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:03:17+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_de.pkl",
+    "source_sha256": "5a5409fa507bee4919b9dc2bd6041332060af7359d785da31e3fde15ce5320ba",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ee.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e6be9324e707669c6a7b9bc32b930601b432b1f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ee.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "65f37f08fd3b3dcb8ec8e6a5d75d7c3c15ff6fc577617699e8cd73b73a2ff1c8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:59:15+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ee.pkl",
+    "source_sha256": "553ddd36cf7f7e8675508b58d0e0e5dce44ab075562b676c1b9571d97229781f",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_fi.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a4d7ce50f0a7be83f8c5506c65abb8eeed4a76b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_fi.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5160eeb41f8eefcf98d31b52c58de287ed50aa569aee1e3613cf97787b3c29b6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:59:58+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_fi.pkl",
+    "source_sha256": "df785a41be2d9335b194ada31f781f376abcdf23caefb9f409c997482bdeb0fd",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_gr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..64a585b9a748495a03fc024d9397472b4e80e211
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_gr.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0313d641c2392ce122960ba29644a400ec060b3fb7867ab323c4d4f5a10fed6e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:12:06+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_gr.pkl",
+    "source_sha256": "6eed1420418535660374a3710104644c1db5f936b53c28d17f7da7780039d285",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_it.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b595b1da54e851faa028d8ec58586b168042cd3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_it.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "92c3df3294cd013a10c5a08bbdc2bc986ff17528df477104b8436369c41ee4f5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:15:35+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_it.pkl",
+    "source_sha256": "5fe9f517a0261be3ad44e57c9cc2a2a1f50684f986ba06eabd8cc5837ea00813",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ko.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..248baa56988e2c3cca1a35ec8a21c1ccba48bc4e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_ko.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "024803456223eb0393727482563ee7af382515afd2a574003aca854cf2117da4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:25:05+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ko.pkl",
+    "source_sha256": "b22407d33038a4f2a9648e70937a59ae8a73322ceff56720a3bb55ff19eeb266",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d23973f654b1fe206de36b4575fd578d5e0c110
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ba.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d6e3dfa7c060c662f94f9dd518b7442cb76ec5295e3f6646e7653afe44608002",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:17:34+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ba.pkl",
+    "source_sha256": "d441b0f8277715db8fe7411a42f2339b873fbd97ac4115f7567b17b55d2db2f2",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..33ffd7be11a6de07b4b65286f879e0eb419d5ffe
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_bp.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "092430ab01129ef5d35b1be3ce31c8b9503b6c8563c6979140efde6cf697a4c7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:20:14+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_bp.pkl",
+    "source_sha256": "40c3df003542601ed91835030e38c25dea00920fb1b1f9bf268934059685f525",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..02408f4a3aab753a035d68acd0b9b7f92e45e1c8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_po.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3fca51454ea4bd7049a7d488faeee30de7c05f4416391e13cc4b6e23ec0efd32",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:13:16+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_po.pkl",
+    "source_sha256": "c085f0967f4baff3db5e3553eb9cf7836a4d10ed2a3a4e568c59a9339c32c62c",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b88b6965ae39bc3830b58277a929f17b88d160e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ge_zu.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "788095b5b58d30fc281ee78edfe025a7f336fe7d32b783e279b4afd2aa202b34",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:12:08+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_zu.pkl",
+    "source_sha256": "df0049856d52a2f3419ca06b0dae19565c9bd248eb63157597f3462d91c68eec",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..0bcfefa09eaeb1271538c1ca6647658be9c03532
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_ic.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7f2bcde31de28a1d3ad7fdeac4e19090599e63b4a08ac09a777e3ecce66e378f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:47:05+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ic.pkl",
+    "source_sha256": "851a3e197de7d056981b825c33d7e08085144d3212b100c00862d274a667f485",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..71d6cf7943adcf65286224ebe9100cfb890a3535
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l1_w2_tr.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d76cc81c87be60484299ac0ed20223e792f0ef609993043addedd136743033ac",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:58:50+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_tr.pkl",
+    "source_sha256": "1c4b39cc5545623bd521e306a118af44f8e482a01c1935293dc9be99f352cfbb",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l2_w2.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..320b3a6eadb1ced418a04a5d79e9aa56c976a82e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_l2_w2.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2494ba163e617c6967c3a06f367cb7dabf49f97901d20bc80764ea74a189917e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:01:00+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.pkl",
+    "source_sha256": "f8843ceb19b86e037775bb320a55632ffd2068b4b89aae19bb7d748329093208",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_tr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c5604d1805b213925d4c6b8cdc1371d026688ce
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_meco_tr.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a9e7bedbf2aa1fbe962083cddfee2d7b949967593bed2584ea654d9844c0f45b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:01:51+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_tr.pkl",
+    "source_sha256": "aabde1e12f9cf1d25553a1d6a7c3c72d8c271707a4a46c810962788104e5b853",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_onestop.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f2ebb447bb58f2cf478d2e14fc1d1b33b0c33b8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_onestop.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d2ca9cf0e52caac12428e8a5b919b126cd9c4dc5edc26afa8cdaf9fc8f77bbe4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T13:06:20+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_onestop.pkl",
+    "source_sha256": "b7f22ce0c336db02563b16c04ec2e0e1fdc76e658fdaa462213afa4420fb9859",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_potec.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..0053aa0ada4a457e15215e14d7dad82025f08872
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_potec.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "101e60ad2c5eb86de816a529b43bbc9096c8a86bda0186c4df28c48600eba3b0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:25:41+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_potec.pkl",
+    "source_sha256": "91d72654cc4135472e183aa40848412f7f7e3d92935f2a4cff6d010780e731a5",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_provo.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa95b463f908ca96221f3284990af13e47b8be41
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_provo.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4e730f74710682ca1e875e20156e8f9a3cf4f8e8ef8e3ea5a262a687aef299cc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:33:29+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_provo.pkl",
+    "source_sha256": "c0f53d56d0852245e0c835afeca2e3a61e4488aae4d1edb6a4515b4a70872f41",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d539554693deed6085e31ebd9e5237d1ce53dbd
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9192f624c5172fd6d756ac6dcb1caf4b903377c2a5b97bd636f41857e2651104",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:52:08+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_sbsat.pkl",
+    "source_sha256": "83e28b16d5f401952701382414954ce9bd5b9b2578e1f1290a912f87aaa37516",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_story_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8e93de1d75cbe991ba314f4c9b354cd666654ce
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_story_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d64a87d36b8f3a6e21b91efaac561bb650d21347f8635ee9481f6b18d5bfc188",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T01:18:43+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_story_sbsat.pkl",
+    "source_sha256": "dff14136df492f249e4f27543f9161c97311c79f80e956a20edb31fa2045ddf3",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_celer.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..552ae3c76018462e2c175a4d384bae4d474d7f5e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_celer.json
@@ -0,0 +1,72 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "9a8c9a95b7231e36b6cfb33e3ed9e86c022f56200a900603a3f42f5f27a01b12",
+    "n_stimuli": 28535,
+    "shards": 16
+  },
+  "status": "completed",
+  "payload_sha256": "dd3873358a62bc583baf6412f8ad1c573ebcc85acd0e33bd14aa886d0e5c380f",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX, Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T13:36:06+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_copco.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c50b8f7ed60a13df098d233779f9a1d3275fc50
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_copco.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d802b1d294523c261e1eaaaf32a63e9bbd687ff9ae6089c0d430b3c182d35924",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T16:25:33+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_dundee.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..c16a5df1709c4596878ec6a517a9f48859060a37
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_dundee.json
@@ -0,0 +1,72 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "55e4d5b6adb8f258a2d2e10aa3b9c8ba3cc43942bba3df2f1fe2ec337793c263",
+    "n_stimuli": 20,
+    "shards": 20
+  },
+  "status": "completed",
+  "payload_sha256": "2132af180933f2f7d3edcad062cc376ca1879f6330ce4a39700d532bdd4b3401",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S, Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T12:40:14+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..208f601ce65d707e512eb62d17c952efff64a844
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "43ef1e386fb8e346bc7c930d93e2b4d21c38b1275dde8ad02a9596c97549c45e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:34:35+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_geco.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..b44a149f97c081900db254b078e96228e85de603
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_geco.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0a1da831d7db529123b3f2c5cc1ea12b1f1b4a3a49e50b8c57b1603d90880fe8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T04:30:15+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_de.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ccc2fa23f3d9622af12e55c39583176483f895f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_de.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5a5409fa507bee4919b9dc2bd6041332060af7359d785da31e3fde15ce5320ba",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:03:16+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ee.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a4bc01fd1f3721b99643ce0a99c6687c401f037
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ee.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "553ddd36cf7f7e8675508b58d0e0e5dce44ab075562b676c1b9571d97229781f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:59:14+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_fi.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8e2bfddca87def1d6650098b1c4bfc2192ce860
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_fi.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "df785a41be2d9335b194ada31f781f376abcdf23caefb9f409c997482bdeb0fd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:59:57+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_gr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..99e803f3a02cc6aa8ddff083f3f138c2f40a2b97
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_gr.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6eed1420418535660374a3710104644c1db5f936b53c28d17f7da7780039d285",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:12:05+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_it.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c80073cbbc6b55047993db83eb009dc4c999327
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_it.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5fe9f517a0261be3ad44e57c9cc2a2a1f50684f986ba06eabd8cc5837ea00813",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:15:34+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ko.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..f53bcc71690d6cde57f2adec3f75b35d1466fc0d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_ko.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b22407d33038a4f2a9648e70937a59ae8a73322ceff56720a3bb55ff19eeb266",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:25:03+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..da0b49900988504df9ac96b9de20fbc5d5c1755a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ba.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d441b0f8277715db8fe7411a42f2339b873fbd97ac4115f7567b17b55d2db2f2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:17:33+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..d64c57094cc913f8e7affce0a121e330ba689218
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_bp.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "40c3df003542601ed91835030e38c25dea00920fb1b1f9bf268934059685f525",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:20:13+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_s.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_s.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c8f9a14955bffff8a6514796a8530f213550cd9
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_s.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9f27685ea775f0843723c5aa8c5b226b7d3420ccb213afd1eb8198439dcdb6a3",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0dba2f1200426cada56b12ee94af96bd41ebfbf19f8524518160c087890dc67f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:32:09+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_t.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_t.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3368675e45183dcd27601bfefec929df7f902a7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ch_t.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "866884887c127db9b93a7cf7e4deb154d4c2cba960aa58ede0d05493044e3877",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0115e98c3224485349800cece54d5ce39b5cb83243c9c62889b8995cc268ea23",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:30:39+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..4122ca5ebfe4e6ce0dc030f97afe9b87493c4300
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_po.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c085f0967f4baff3db5e3553eb9cf7836a4d10ed2a3a4e568c59a9339c32c62c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:13:15+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..6754d73f2ecfee7a3e2b9b9829c6c1b7e888b582
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ge_zu.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "df0049856d52a2f3419ca06b0dae19565c9bd248eb63157597f3462d91c68eec",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:12:07+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..122931f730526c034c607dfc8b13937c4ca7c659
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_ic.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "851a3e197de7d056981b825c33d7e08085144d3212b100c00862d274a667f485",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:47:04+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..69ee0764f92ae0fefb0b4ca6bab96c9796b5fc9f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l1_w2_tr.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1c4b39cc5545623bd521e306a118af44f8e482a01c1935293dc9be99f352cfbb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:58:48+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b1200ba08b814b6a1845879844b3ade832d65c0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f8843ceb19b86e037775bb320a55632ffd2068b4b89aae19bb7d748329093208",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:00:59+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_tr.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..34bdce55880dd13d4e2d9f8d3517e4b80a3de82f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_meco_tr.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "aabde1e12f9cf1d25553a1d6a7c3c72d8c271707a4a46c810962788104e5b853",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:01:50+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_onestop.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d6ebb15263b4f5edaff608dd0e108340910f5c5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_onestop.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b7f22ce0c336db02563b16c04ec2e0e1fdc76e658fdaa462213afa4420fb9859",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T13:06:15+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_potec.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..63f5cc8e3d6f715bf09592440b87df964fbf7def
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_potec.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "91d72654cc4135472e183aa40848412f7f7e3d92935f2a4cff6d010780e731a5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:25:40+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_provo.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..49ff584bf76542fc39ca318966ec096df99eec70
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_provo.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c0f53d56d0852245e0c835afeca2e3a61e4488aae4d1edb6a4515b4a70872f41",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:33:27+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..40c39da872dfe357a438a383210ece877257fd48
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_sbsat.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "83e28b16d5f401952701382414954ce9bd5b9b2578e1f1290a912f87aaa37516",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:52:07+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_story_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..d941a7191b1a0d2cdf0d06cce647fd863109fa16
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_story_sbsat.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "dff14136df492f249e4f27543f9161c97311c79f80e956a20edb31fa2045ddf3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T01:18:42+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..29ef8913fe546198a8783eee79abf3d0bb865f77
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "48d5662da3ed1500a648a8b983848c3f28b2b291a236b53dbe132d4de90cf9ae",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:45:04+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 251,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..fafe4b1412cd3c7fc950742f0a2bdb2d0720a0ed
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "19d949b0b9c1f4f2093795dc7ca180afd4ffb92a53338fb4edb968f6e1ca98fa",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:45:06+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.pkl",
+    "source_sha256": "48d5662da3ed1500a648a8b983848c3f28b2b291a236b53dbe132d4de90cf9ae",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_celer.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..2548d6ed1e2bd2447d07ba2187352da07153b265
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_celer.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "dc628513fbd6c8191f5fced44e013b5ce6eed5e1760eb2df96277ef5005d965c",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T02:23:02+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_celer.pkl",
+    "source_sha256": "ba50298cad4f40b1e1d60eb269389eabf8461fa2bb214882e9b5b72ab1c4ad77",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_dundee.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..4239ba99a8378d61638e7b4c87922713fd5d9744
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_dundee.json
@@ -0,0 +1,72 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "55e4d5b6adb8f258a2d2e10aa3b9c8ba3cc43942bba3df2f1fe2ec337793c263",
+    "n_stimuli": 20,
+    "shards": 20
+  },
+  "status": "completed",
+  "payload_sha256": "769539eabe661d5f2a3ad08b3da761343c00f5c4a142229f2b9c85ae7f316009",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-40GB, NVIDIA L40S, Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T08:36:47+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_geco.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..c88103c82a078ed811074e21e88c24a906724150
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_geco.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8c9f0d58ce2cd9f3666d3de71eb8b8198b083c6e141355908794378be8474f1b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T23:17:32+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_geco.pkl",
+    "source_sha256": "e5ca5c25cf66aa272d4dd501c02d85ae5ad138d69884d369f60d3301df20755b",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_de.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..a34bc1e36cd2396dae04ce46e004c09bdc5c63f2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_de.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "96d6e71e7531a6381103c91cd95f22b2240b685973ec7378e3b92577d98ef715",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:11:04+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_de.pkl",
+    "source_sha256": "7c0906b5e4af7d41ab606c83ba12bf57266fa9c8c190df6fa1b5ad3109d0450a",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ee.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..9602b1a99f6cd724d1cb0532442d3bee20820d9a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ee.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7afca447977b942961419d0c4e318decd7140dbeb73baa6bf21440422595e3cb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:35:10+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ee.pkl",
+    "source_sha256": "6ac6275ec30d58f8f6177262447224a4e23dc6bd60f2bd227cd070cd83ac4ee9",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_fi.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e50b8d90c92a3d2f80a88d5dfaa83e08cf9cac7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_fi.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "80e591b6239926f6a6109c9790f4d38c256c487a8925b7c936ae383141e3be22",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:24:45+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_fi.pkl",
+    "source_sha256": "279f9da41d2fd11c07940aa60a4768d698f420cc083fcbea76efaf9103ae86f3",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_gr.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3a62183a494e41b11d90edfca9c7e50d1446659
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_gr.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cf488467e81d90e56476ed927ebc0c66cd419bbed634172b2b12ef4956397a65",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:41:00+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_gr.pkl",
+    "source_sha256": "3db7f74f0304110a8f58ad2118052c0f7ad81b49020d8d46e08c2504c64a98ac",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_it.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..9ffe04d01dfdca96755f3009c6696c315837a723
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_it.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "717432cbc805dfe70651ea98822d3aa08e22deca320e70a127e590b486d1cc1c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:14:19+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_it.pkl",
+    "source_sha256": "38fe679a31a6653ae67c782659f428680ac41c0b4efc92e446a7ed104541c6f4",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ko.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..d876691b20151d61658a91096ca9f8045d14f73a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_ko.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "26362b4ba571ca95b78de2e29f6de3e088e3fc91800dfdf9e6c3a9dbf09cfb01",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:37:55+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ko.pkl",
+    "source_sha256": "13365be5539f21e06a74988b4cd88ebf16973b989d7f6a319f9226c0bf7fb01d",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_tr.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8a8a5accc96c87682600c8709a8c0cbdc2db0fd
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_meco_tr.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ea01900224f0f337f147a25cb8b5957853eb1922c5431a15296af4074fdde66d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:25:57+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_tr.pkl",
+    "source_sha256": "cb54b33f8eae95c327342c359b05bf953b34b242a9e305d3d63e37f099998598",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_potec.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5090a683374e76165c024c1a1028a3f415c858d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_potec.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6943051d05579914f99dc7c9191d7f512c5857e185edee6b4135038a1e5e97a7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:20:13+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_potec.pkl",
+    "source_sha256": "fbe6355c21ce8122ffe2015796cd44e0d8ae622df592d982b2c7bddf3d33aa64",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_provo.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4ba7366a468a755bdac80c943efbc73c7c16cc
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_provo.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "274afedb17700862d3cc3e4edd03e8a91c509a80f00d7a6d5cb8180b1b178c24",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:39:04+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_provo.pkl",
+    "source_sha256": "d3e0168a4580583275a74f56a3ee95c6cf611e6bf2f7dff35b354c546a91d488",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..b414a9bcdac4feb364171a021d1b2e9294c32bdf
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9927c6977aaa0904e407d8331fa942ed1d90321b924587c12d69548426749e53",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:13:15+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_sbsat.pkl",
+    "source_sha256": "448193927898b115dd6161b0aec2795e12c3c8e723bca454e5da8bc8f227f644",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_story_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..147b516452729d59731a399830f11a25e17528ce
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_story_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "33307728357b43bc08c49b0fdc8a4c16e38a6ff5ddad1d4b4d127ab8f8076c52",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:15:34+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_story_sbsat.pkl",
+    "source_sha256": "8ea2e12914bb12a4217448c2e9b9abf95cd720057176c1c04aa75bc155b8398c",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_celer.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..25c86261de40fc91498d70c6ee80098278770789
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_celer.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ba50298cad4f40b1e1d60eb269389eabf8461fa2bb214882e9b5b72ab1c4ad77",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T02:22:42+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_dundee.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..b7a3d124d4d63e52c8a48590bd098e51a4b4eda0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_dundee.json
@@ -0,0 +1,72 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "55e4d5b6adb8f258a2d2e10aa3b9c8ba3cc43942bba3df2f1fe2ec337793c263",
+    "n_stimuli": 20,
+    "shards": 20
+  },
+  "status": "completed",
+  "payload_sha256": "b520c4eeeda0684f8cdee59469c5e29973a2d8c72d88fbe62a1ade876cf9a4f7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-40GB, NVIDIA L40S, Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T08:36:46+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_geco.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..402c0834ff71f8fd8c310fb0853dd45fca00c327
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_geco.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e5ca5c25cf66aa272d4dd501c02d85ae5ad138d69884d369f60d3301df20755b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T23:17:27+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_de.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..24cc3dacaeaf0691d9fd0b8eedf42d3fd3dc4e74
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_de.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7c0906b5e4af7d41ab606c83ba12bf57266fa9c8c190df6fa1b5ad3109d0450a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:11:03+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ee.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..042400f4436b5cf96334497f61f372a5efaa5e49
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ee.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6ac6275ec30d58f8f6177262447224a4e23dc6bd60f2bd227cd070cd83ac4ee9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:35:09+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_fi.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..3323af79cc9880e7984bbace8db3aedbe7d5f118
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_fi.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "279f9da41d2fd11c07940aa60a4768d698f420cc083fcbea76efaf9103ae86f3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:24:42+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_gr.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..0d17baaecc56d46734247c08b5c60b35ca75d550
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_gr.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3db7f74f0304110a8f58ad2118052c0f7ad81b49020d8d46e08c2504c64a98ac",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:40:58+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_it.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..c69419d0f197f16c4c9f64ba13c3944cc28c63b9
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_it.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "38fe679a31a6653ae67c782659f428680ac41c0b4efc92e446a7ed104541c6f4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:14:18+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ko.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..07314c245d6fe2b915c71c0a0bff44c963a597b6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_ko.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "13365be5539f21e06a74988b4cd88ebf16973b989d7f6a319f9226c0bf7fb01d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:37:54+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_tr.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4d129a882df745a4b0e8de824b70e73548aabae
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_meco_tr.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cb54b33f8eae95c327342c359b05bf953b34b242a9e305d3d63e37f099998598",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:25:56+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_potec.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..cc65ed2b5a9f11e2399f642e23b34f95859b1c12
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_potec.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fbe6355c21ce8122ffe2015796cd44e0d8ae622df592d982b2c7bddf3d33aa64",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:20:12+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_provo.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c226e3e03120f8a4d281b6ebb19b8ed21a49443
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_provo.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d3e0168a4580583275a74f56a3ee95c6cf611e6bf2f7dff35b354c546a91d488",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:38:57+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..b20cffa4955343e37343e4e13623c8648ec079c6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_sbsat.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "448193927898b115dd6161b0aec2795e12c3c8e723bca454e5da8bc8f227f644",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:13:14+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_story_sbsat.json b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..18811d7f8c80a470f12265d5cbe3783fe1e4988a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_cont_entropy_N64_L10_s0_tokens_story_sbsat.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8ea2e12914bb12a4217448c2e9b9abf95cd720057176c1c04aa75bc155b8398c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:15:33+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 512,
+    "half_size": 246,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_celer.json b/manifests/ai-forever_mGPT_ft_entropy_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..9751c81111ae6c34862ec71fbb85438824884c0e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_celer.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "de8ebec4bea4828d1949046310a84d79627a83f77379b7a98ea67328bcd0c093",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:23:50+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_celer.pkl",
+    "source_sha256": "88d2e408837aa90e8011b8f45063262c0486c95a322b8a19688b19fbfbe6ae79",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_copco.json b/manifests/ai-forever_mGPT_ft_entropy_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9ae2adb328c117b5bafb29063b444e46236503a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_copco.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "24c1f8dc5100389d46f9a2f4a53f2bee1d0cd0e48937f2c74caae0e634f942f6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:20+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_copco.pkl",
+    "source_sha256": "c771853e503d118ab1768948528de1d4333968295341cf90252d9d603636768f",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_dundee.json b/manifests/ai-forever_mGPT_ft_entropy_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..a12cd0ccd79700e812a9bf976f6ca684cf6dfb28
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_dundee.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b4092d0e3f8738d3c8f5d653fe7d3459994253e85206206254222da61cbd15ca",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:20+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_dundee.pkl",
+    "source_sha256": "fa3603fd3311da97eacf68ba8ad98f7aa9c56f2a4e23234c78e3a767a97ed842",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_federmeier2007_n400.json b/manifests/ai-forever_mGPT_ft_entropy_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..d77b25f1fcf82db8d2c15a073da68e7d366fa019
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_federmeier2007_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1a2369b0cde3a3463a122796ee9b59b47325886d36bcc8dadd66cbe357751968",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:56+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_federmeier2007_n400.pkl",
+    "source_sha256": "93a16674accb2579215cb8c9f0be416399e9263d7e39663259fe1a155c81ef20",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_geco.json b/manifests/ai-forever_mGPT_ft_entropy_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..36edcb94af7c730ff401f0aa43a321f83624e2a2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_geco.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "399231ea138d094684d43b5f73806fcf36563f94e95f5b458cf15e825d4b980c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:32+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_geco.pkl",
+    "source_sha256": "77dac7ff6894d93b7651d4eaa93b678899b8fbf796fe507713f719dfc0f69f04",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_de.json b/manifests/ai-forever_mGPT_ft_entropy_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..dadc349fe02589c7a56d91ebd304e0a5680e2684
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_de.json
@@ -0,0 +1,66 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": true,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ]
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "ca57e94049031d6937ec92f01c668130b26b03ef357c33b5b753f91a783d6626",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "588bc1c1bcc98e7864e547ce053a679e32b7274ac620213e5dc4a796c401c01c",
+  "environment": {
+    "code_commit": "dab4617575e68ca1825e76854a713f747f1900bb",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-23T13:53:29+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_ee.json b/manifests/ai-forever_mGPT_ft_entropy_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b5fbf7ae4eaf06cd55efef834efdd1cf70bb256
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_ee.json
@@ -0,0 +1,66 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": true,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ]
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "ca57e94049031d6937ec92f01c668130b26b03ef357c33b5b753f91a783d6626",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "866857333041db27c01ad30c58881a11be73b1ce53224eb34d07d71c17eeab80",
+  "environment": {
+    "code_commit": "dab4617575e68ca1825e76854a713f747f1900bb",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-23T13:48:59+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_fi.json b/manifests/ai-forever_mGPT_ft_entropy_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..30ac1a04f68a65de0974ac96d47a12b210b60c16
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_fi.json
@@ -0,0 +1,66 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": true,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ]
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "ca57e94049031d6937ec92f01c668130b26b03ef357c33b5b753f91a783d6626",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "77e8feee4342e68d9508da226e5d7042e0ff313ac9fa607b7289895d51d28bad",
+  "environment": {
+    "code_commit": "dab4617575e68ca1825e76854a713f747f1900bb",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-23T13:50:30+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_gr.json b/manifests/ai-forever_mGPT_ft_entropy_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..091db62f3e42649e5fe129895187e89b306818ce
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_gr.json
@@ -0,0 +1,66 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": true,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ]
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "ca57e94049031d6937ec92f01c668130b26b03ef357c33b5b753f91a783d6626",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cd83d7394f478256c74303d852dcbebedb6ff3d2c55d977ba927c3c250da6693",
+  "environment": {
+    "code_commit": "dab4617575e68ca1825e76854a713f747f1900bb",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-23T13:55:08+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_it.json b/manifests/ai-forever_mGPT_ft_entropy_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a4b2e57ae756ecb390f72f9ac73dc1d8823d13c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_it.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ec5a803415bbcef69b47e11269841af62b028aaefa0c0a199b35ee7dd6b90601",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:28+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_it.pkl",
+    "source_sha256": "3be1b2039aa69a102cbdff274ff00823d20f22736fbca4b21f84cb20855f59e6",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_ko.json b/manifests/ai-forever_mGPT_ft_entropy_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..4bef32782fe9e7e77131b9476efa41b154aa5f13
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_ko.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "800f1fa19b6a752e47329cf462d333c52eb0eec1b6cf9a69b95569504903b106",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:31+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_ko.pkl",
+    "source_sha256": "41b2ae3080ae2d75a2e848492fd6886330d5c8cabfad1e07300ae79be3b7df96",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..df33f18b79db459363ecbb0926ddeb7d044aec61
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ba.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e44e7b06021a5f3e98614ba3b491a64af971ee71d8f3680989963d1fbc856c56",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:32+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ba.pkl",
+    "source_sha256": "def3e00e0d8691579764748d398eb2e9eced0fe815d444120e78ad0ff6233464",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..704486d1529f364416109112212f8f1a669a742e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_bp.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2c665dd9dfa5b23af71093caeb03419b6e875b01eda44455b709a892af0ce017",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:38+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_bp.pkl",
+    "source_sha256": "b15c4a91d2bbd8c59c04acd174b048c01438a05cbecd23ac8e287d0550a92926",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..9dddb8b986ab49a48e9f7982c22c0f5b3a7015a2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ge_po.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a1d660d74ce439e0ac68bf7aca1f9e1fe6d78372167025307855e9c9bde7f324",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:36+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ge_po.pkl",
+    "source_sha256": "6bea1ab5c946d3a3eaaabe44eb23d9e2319a18aef473bf74221fb5accf78e8c6",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..a87286ff57c84d415abe9daada2db77c53c1a28f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ge_zu.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9684d19c64e25d499bca7799869b15e648022308133094449e523448869b1652",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:38+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ge_zu.pkl",
+    "source_sha256": "581afab4aac0701b67e960433a68af1b6cc8ad2d0d1fec6ee6c56b3023d57b9b",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..d08bd9d4cb3015d64ef572d32d7a66f561348dfe
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3187ef53f7953cf125c38efa8cb7ce25fe364ce200b55e0b7d8bf35523ed22c8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:46+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_hi_iiith.pkl",
+    "source_sha256": "07426ee31b1e146baff1f474446594e312b2280462cb267e8756627c27cadbfe",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..046444ff5e50fe0348baceaf3f0ac0fecae7db81
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8bcc92e37fcd87af519ddc31dd7806ea34c2d833fff29b6512c8f019c9626e34",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:49+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_hi_iitk.pkl",
+    "source_sha256": "a42a27fbc51be94abab90dd7fa53920bcf32434acacff9ac261851655ee650bf",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e92093fe07ff54bf76c7516eba4ad378ee541cb
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_ic.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e0a0c363b4e419b2c46069fa6444a2d6587268bcec4602a9f2723cfb881278a7",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:14+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ic.pkl",
+    "source_sha256": "551ab0ed82c11a0b2b0028fc5b2af0261a2e6287f1912f2556c33dc212fde5a7",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..60e3ca308a6bd07e655660833a0456a4b91bd0d8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_l1_w2_tr.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9c1893610735a28f5b570a6b36d22ed985667ff2bf7a357f7b352a706acf8a95",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:14+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_tr.pkl",
+    "source_sha256": "d80497ca9ef8599d7ba1514db5ea5cf95cdd04688ae1c8b43cb1a0054b0c789a",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_l2_w2.json b/manifests/ai-forever_mGPT_ft_entropy_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..f319531a4751da60fd0a5a37fc316f88981b0a1b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_l2_w2.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0d61e4ba2c1513844e52c8fd1b0665376d6e2d8097f2e7da6327ae74471a48a4",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:42+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_l2_w2.pkl",
+    "source_sha256": "26b6d94b52f55da31407e56b65c61d16028b6e837c6d85dbb493d95293749b98",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_meco_tr.json b/manifests/ai-forever_mGPT_ft_entropy_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..7179045c3c4d32449a45ec03f66bcbfd1b414dc2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_meco_tr.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "64431527c87a28f17ec87f83942f5d60823faed3f2a9ad64b917d86fbcde3272",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:46+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_meco_tr.pkl",
+    "source_sha256": "431a641b183f8585aec8f8dc32e85f6c5bc912c0e91a387e73feb32b9ba2000f",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_natural_stories.json b/manifests/ai-forever_mGPT_ft_entropy_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e0d5f330b9de801f6318505eec1eaa5272117b3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_natural_stories.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "478bd93b093325d46f54f292f5884317078610ca96e4e8e728b415f0783ee02c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:54+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_natural_stories.pkl",
+    "source_sha256": "b7e50eb1df19cc34180749854ac8ebb2a70344ea5b35a934d0d6577bc48a9f44",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_natural_stories_maze.json b/manifests/ai-forever_mGPT_ft_entropy_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..7de4d107949c3b36bae76dd6537c731850901f1f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_natural_stories_maze.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "544e97f97a9c07638b453835e42e9ee790c79a03a242e1b6106955aa3e20d976",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:05+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_natural_stories_maze.pkl",
+    "source_sha256": "16c5289ca4caddde8d24f081e5b558938100147050065ac0a18d7e8c7f4b0025",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_onestop.json b/manifests/ai-forever_mGPT_ft_entropy_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..0fc76663928b76fcd7a80594fa67d5c8dca7f5ac
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_onestop.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "89141adcff931a8f9c80ba85999dc9772d51d87ee48c0e750b8152f3545fd527",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:47+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_onestop.pkl",
+    "source_sha256": "0644c13d4e509603d06858891b7689f21a4824f6a47e14f6fcfef97871d956c6",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_potec.json b/manifests/ai-forever_mGPT_ft_entropy_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..82b63754fcf328feb0b39cc89607fe78d9ae23ed
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_potec.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "79ef7ec6e435f36984afd8c1a643e4b14130fa90fd4479875214200213d20b4d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:11+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_potec.pkl",
+    "source_sha256": "3471dfc9c00fe9ab9196f1e72ae0e677e6b4d0c190c9d5ad24676b237cbdcb94",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_provo.json b/manifests/ai-forever_mGPT_ft_entropy_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..fab9fd0c83223827d4799aee66429a60e487e452
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_provo.json
@@ -0,0 +1,65 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": true,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a26f8bdcbce9406727b14e381d6d8115eac1d1fce555c372f2b6018f648a357e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T19:08:55+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_sbsat.json b/manifests/ai-forever_mGPT_ft_entropy_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..741f2e532db7aed669483f4c915e45f6e1e39ace
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_sbsat.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "170b12c962318ebc245cadc54494f8edadb27288d26f509f4bf01d72f2a98338",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:28:01+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_sbsat.pkl",
+    "source_sha256": "4e1ceebcc6650d48480d82a69703129ee29c1a6eb38a0185abed732142a2a628",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_story_sbsat.json b/manifests/ai-forever_mGPT_ft_entropy_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..22031bdf7770e94a1a9d883172e51e7fd2af9f1e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_story_sbsat.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "41c4bd8d106f0b713bf5c660e98cd5a558af32a6b318e0f972da30cf0a4ade9c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:02+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_story_sbsat.pkl",
+    "source_sha256": "b517b56855c7e08d177de62d5d345ee7f85720af3a43b56421a4e37eba40e41f",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_ft_entropy_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..a49b343856985f4dfac4ffbf2ec4b590ead4bef5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4b74c68c6b637bc377f7b2698da46d80ea049df26eb89762aaf0b55cda693b0b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:10+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_szewczyk_federmeier2022_n400.pkl",
+    "source_sha256": "bba0d05654b8fd7417c1d82aae287140bf66d652f77d1906d18baa26e900838e",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_szewczyk_n400.json b/manifests/ai-forever_mGPT_ft_entropy_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..777ad827c7c18596ac4313150429464e23838d8f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_szewczyk_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f804f937f287ca78d1c6ef99017268a48fa8bbe5bb3c562fe46028bbac7a578a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:37+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_szewczyk_n400.pkl",
+    "source_sha256": "e40af2f7eb654112c584dcd72c4dfb6868093f6a37b344985e0b7b0dab5c912e",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_celer.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a6855758a175992cde3e20acb58fd7191d25636
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_celer.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "88d2e408837aa90e8011b8f45063262c0486c95a322b8a19688b19fbfbe6ae79",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:23:32+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_copco.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..f04112458f4e4f5adefc2292abeb23f390831cbe
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_copco.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c771853e503d118ab1768948528de1d4333968295341cf90252d9d603636768f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:11+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_dundee.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..0bf7e4c2f564036ae662ecbf5334def043c0e5e3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_dundee.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fa3603fd3311da97eacf68ba8ad98f7aa9c56f2a4e23234c78e3a767a97ed842",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_federmeier2007_n400.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef0cac4f9835170e9958f21dd11ce71cdfee9ea6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_federmeier2007_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "93a16674accb2579215cb8c9f0be416399e9263d7e39663259fe1a155c81ef20",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:48+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_geco.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c65d3c4b376803308b8d942de123365f832d4f0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_geco.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "77dac7ff6894d93b7651d4eaa93b678899b8fbf796fe507713f719dfc0f69f04",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:24+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_de.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..2df12f5786e8f194dd749a3acbf3f43ee0868cd4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_de.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0f7c0995e2e0a5a4143dc6809cce01f97ec4d505efe47eeb959f7a4a94bd9a93",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:28+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_ee.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..b335bab5f71b60b8444df7b6caa89ccf376b1cc1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_ee.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2d937485f22783e55197053dbef4b29f1e7d30f15f2b522262baee164659f3a0",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:03+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_fi.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..f181a9506eadde9b341ac62c724a81d4f2263408
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_fi.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0e83f11319826bf7a94d93f8a04c176f3c41468d018c8588182c47e35eae4b7a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:06+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_gr.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..79e2203a49c01c260eb6058f5dcb35c1bb2f3497
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_gr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "31f49efc245a68790ea581b9c0c97d2dd84f69ed561954924220a4f60b656a29",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_it.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..1adcc4afdfa8b121af721884e1c5f017d1804475
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_it.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3be1b2039aa69a102cbdff274ff00823d20f22736fbca4b21f84cb20855f59e6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:22+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_ko.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4998c81b3b42a2cbb890182564c2b8ca2a72892
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_ko.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "41b2ae3080ae2d75a2e848492fd6886330d5c8cabfad1e07300ae79be3b7df96",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:30+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec79bd68b78cbf0300c7d696c6016ea926882f28
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ba.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "def3e00e0d8691579764748d398eb2e9eced0fe815d444120e78ad0ff6233464",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:31+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..15b412380411e6d0ef1a79ee0e11b105b1216b0d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_bp.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b15c4a91d2bbd8c59c04acd174b048c01438a05cbecd23ac8e287d0550a92926",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:38+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ch_s.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ch_s.json
new file mode 100644
index 0000000000000000000000000000000000000000..9337dcec167cb4d24acf35325f12b2b8cd6a60eb
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ch_s.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9f27685ea775f0843723c5aa8c5b226b7d3420ccb213afd1eb8198439dcdb6a3",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "28c8b2c2759a85f5a2b7b796848abffc8b313f6920797389e83fb1ee9b33a53d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:00+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ch_t.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ch_t.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd43340106d6cd54b3a2192e5b5e65b5e7345fc9
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ch_t.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "866884887c127db9b93a7cf7e4deb154d4c2cba960aa58ede0d05493044e3877",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d5de038b5270331b8d67462f0d8bcb60c8b0be145edb0a367f00ca7dadfe6d98",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:07+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..21089a1850a29d385c9682e82785092c092c6a85
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ge_po.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6bea1ab5c946d3a3eaaabe44eb23d9e2319a18aef473bf74221fb5accf78e8c6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:35+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..7774465dcfd61faf854ada54ff6230d578e49643
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ge_zu.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "581afab4aac0701b67e960433a68af1b6cc8ad2d0d1fec6ee6c56b3023d57b9b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:37+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..f21e3e492c62a13c78326e7d07a50cf076a4b817
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "07426ee31b1e146baff1f474446594e312b2280462cb267e8756627c27cadbfe",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:45+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb44fe728f72f91d412a8e6e659c3ad87a0a59a6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a42a27fbc51be94abab90dd7fa53920bcf32434acacff9ac261851655ee650bf",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:48+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..916294e0d99645d4077d57f1e79ef7cd8ce95e08
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_ic.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "551ab0ed82c11a0b2b0028fc5b2af0261a2e6287f1912f2556c33dc212fde5a7",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:07+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..1621f029bdbb6a52bcd8fa911fd36c2598f6e556
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l1_w2_tr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d80497ca9ef8599d7ba1514db5ea5cf95cdd04688ae1c8b43cb1a0054b0c789a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:07+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l2_w2.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..6791695683356c4c434d22478e2f88fd224870f7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_l2_w2.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "26b6d94b52f55da31407e56b65c61d16028b6e837c6d85dbb493d95293749b98",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:41+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_tr.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..98da8e8d64f59db923af4d5bd6cd66bd08e75ede
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_meco_tr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "431a641b183f8585aec8f8dc32e85f6c5bc912c0e91a387e73feb32b9ba2000f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:46+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_natural_stories.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d63bbf9e24971f54030c85a05e3f3b1eb05c687
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_natural_stories.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b7e50eb1df19cc34180749854ac8ebb2a70344ea5b35a934d0d6577bc48a9f44",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:36:53+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_natural_stories_maze.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..e37d2ee939b7a6e7530eb06e573a61032ab5b9cf
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_natural_stories_maze.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "16c5289ca4caddde8d24f081e5b558938100147050065ac0a18d7e8c7f4b0025",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:05+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_onestop.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef60952b95533eea11251e0f188d24b3b040d004
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_onestop.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0644c13d4e509603d06858891b7689f21a4824f6a47e14f6fcfef97871d956c6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:46+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_potec.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..bdabd480e05763681aec051be6e5c48bcf823f37
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_potec.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3471dfc9c00fe9ab9196f1e72ae0e677e6b4d0c190c9d5ad24676b237cbdcb94",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:11+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_provo.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6c5fd9ddc6cdbedb6d6a5b35e0262d55275fcfb
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_provo.json
@@ -0,0 +1,74 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ]
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "ca57e94049031d6937ec92f01c668130b26b03ef357c33b5b753f91a783d6626",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2345071f2385f906337455b48ec02d978da0538caa59a53ae5e84a0ffe930010",
+  "environment": {
+    "code_commit": "dab4617575e68ca1825e76854a713f747f1900bb",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-23T13:37:34+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_sbsat.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..dd081700f71a3ee97834c83eaf4a156b2a7f6579
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_sbsat.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4e1ceebcc6650d48480d82a69703129ee29c1a6eb38a0185abed732142a2a628",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:27:48+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_story_sbsat.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..07ee7ddbee155bdf7bde5f8d0bdafc8026a1fe8f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_story_sbsat.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b517b56855c7e08d177de62d5d345ee7f85720af3a43b56421a4e37eba40e41f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:02+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..0953eac41142e6c1d8741c594db386fd1256ba8c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bba0d05654b8fd7417c1d82aae287140bf66d652f77d1906d18baa26e900838e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:10+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_szewczyk_n400.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e463eeff0bcdfc36bd075fae8fbbedabca361ee
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_szewczyk_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e40af2f7eb654112c584dcd72c4dfb6868093f6a37b344985e0b7b0dab5c912e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:37+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4e820ff949dfd48e835db9ab3bb0910f4babe51
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_wlotko_federmeier2012_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "89fcc450215ee678675900582e44bb4884f331beda5119c4db8f29bf66487fec",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:10+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_tokens_zuco1.json b/manifests/ai-forever_mGPT_ft_entropy_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2b6264375aaa7415a21e6b7be81ef3ca4bb351c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_tokens_zuco1.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "530bd592e664ec69cc10093905724d7636b92c872710c5247cabac0ec602a06d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:41+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_ft_entropy_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..8adb4a026aeb7fe5032707777da2070271e30082
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_wlotko_federmeier2012_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1f56028fa172bffcffe3a2375f7c88f0f4bf2d203bedf9251c1745482d5946f4",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:11+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_wlotko_federmeier2012_n400.pkl",
+    "source_sha256": "89fcc450215ee678675900582e44bb4884f331beda5119c4db8f29bf66487fec",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_ft_entropy_zuco1.json b/manifests/ai-forever_mGPT_ft_entropy_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b2105ac5491dc93c5c27e6efd6671700f5d12e0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_ft_entropy_zuco1.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "091ee211cc2825a70156f385c40e6d111d2040945659624cdf2a662fdee38516",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:41+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_ft_entropy_tokens_zuco1.pkl",
+    "source_sha256": "530bd592e664ec69cc10093905724d7636b92c872710c5247cabac0ec602a06d",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_de.json b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc560369178390d0d0388f8c65a02926cbe31170
--- /dev/null
+++ b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_de.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "22ad49e14e192c48a088aab2ed310a5ff0a62b111e00a62d3413897ad65c8ccc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:22:33+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_ee.json b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..611d541bcf5ab9bb2e7bc39feff93a162fd6a697
--- /dev/null
+++ b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_ee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c86f1c9c382fefa05b0a4011608011990363a415edb64b6528ddc53d394251d5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T20:05:37+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_fi.json b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a0b04d6d7cfd6ee128d063fe190aeebf3d663f5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "776f7dc4e390b427b3f185fcc2e65b0a8cc0c7f62005d8a2fabe09b6b5a9e5ff",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:07:03+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_it.json b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..446436d129f7509a793849abb091864f9613c415
--- /dev/null
+++ b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b07b9568aeb6d4ff38cbd665d157f9403c63cbef13ba3da32ae620b224ff2502",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:10:05+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_tr.json b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..dd9c102803d1f3ceafbfe1ddb52544cfa512601d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e400ae633f0ebeeb1c14d74c839fed1f24cd2ef5761a8d1079f114eeb2497bee",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:51:34+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_potec.json b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b19a3aeba75129cc6230943db94b3f46ba707ab
--- /dev/null
+++ b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_potec.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "96ed2e9b7a726b2c3a0193e51c0469ab24a4bc4c7efc64b17c68fe06b12077a3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:47:42+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.json b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8714d15f714fad6830c0fb7254339b4e12fb20f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "1556ccf78b562771efd140a1ecf61c91e86fc90fce96ffe5e7f0434b0fab4f35",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T22:34:17+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.json b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..17368e88b7963131fccee9d9079348ac717cac49
--- /dev/null
+++ b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2c858990f67aa71d916174bc4bfed195d344b3dd96747fba1aca7c0f2f2c84cc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:03:39+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.json b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca93b451dc7ef3bdf47d7af29cfa8ea58dfefc4d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f6e23335a748bd4b6e9e4dc3164b69582230286dbae6fd97f24744e2d69a3d26",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T00:49:02+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_dundee.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..475a1d5e105a3b3ee9cef09b33706d418da570d1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_dundee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "4b5328e7d8f49efdaf14e3fb4576a9d5e2dc33aa26fd45d1f13a1298d8b1d4ad",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T23:05:55+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_federmeier2007_n400.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..c4cc9af402fdf6235d4b6ac094c4ef4cffc93087
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_federmeier2007_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "30898326b9a5a112edad747a393674edab13b1f4f62e09505ac5e2d1e5314153",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:39:50+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_geco.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..e46516a462b0aaa303701b3aceb51d4fb6a83e53
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_geco.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "5dd07efde8cac880704ba5d991d1f439eea45a201b69604ebede3e7d36499b3c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T22:22:39+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_meco_l2_w2.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..c4e100837427900c81ace89f6a09e3c055a7825d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_meco_l2_w2.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "53e6b410b7626eb35e6ce5f45a719b6c21e9089796a8f7bee839de7ac59fc7c6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:04:04+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_natural_stories.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..d1b831917935b4ad150c087cde1bf30ae035cefc
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_natural_stories.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "d4e7e371691e78c90ce8ef9c2ad50bbf44b33238970051f28e9c70b150a9a644",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:33:58+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_natural_stories_maze.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..b588c9b9b7c2f1ea7d14c8f26d91163f23b616e3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_natural_stories_maze.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "bbbf94f128b12de078fa30be42e78445e3fa8ebecf29365e339bde51c735c45f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:22:05+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_onestop.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..af6932be2a0816b84ef6c03e096c7770dcc9382d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_onestop.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "70047879d58d7effe4a345500fd6a04fc5cdd3b579e4d5d4808be7c1e8adaafd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T22:45:25+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_provo.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..bcfbe822006f24cf29cf3e72a5107e7219c8888d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_provo.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "60fe5bb1453f3425a4dce619d1a65d6e3b8999ffc580216d5857a06cc3c00791",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:06:04+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_sbsat.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..75323bc339122d4f18d00e087ebc06ca45faeffd
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "64d3d9f8778b1ebf4b9f9acb3341593d5582c4611e7d3970775cddba4c53c254",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100 80GB PCIe",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:05:46+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..787b1c286ef63d209efe0aed48ddc5b8cb0e1e2e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "e31f47d8f045057907d242c38cdc2561184f58bdebe2af5e1a6c30abb2f39f75",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T05:16:11+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk_n400.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..1e73c4ba03236cf0d5c79523f2e9505a22fda461
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "c7435d249bf9acc0242d1aa227a9ab92f19a464d4086c7ed3ceff6a06d6e55c6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:18:17+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..0cca5bac605e7f9965fdbe25371fd0f66ee4f3d2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_wlotko_federmeier2012_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "a2b0674a0a8f9989dd85d2d612f53ab5d27b87b58191d872900ba2b3c8688ec3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:57:08+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_zuco1.json b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b2560f411937bed2258bee9f3f716bdd90fef91
--- /dev/null
+++ b/manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_zuco1.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "47c0760a2f9232fc97b695a2a64ffa4f38beb5fe6d930b068688a84f7fd5441f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:09:38+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_celer.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f5e42b1bde491c887472a62bfda0af27a5a9690
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_celer.json
@@ -0,0 +1,70 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words",
+    "corpus_sha256": "53cd9eb062ffb62069b53c2911e7bab7257ab54d1cb6a54b5cdd74304fbdf92d",
+    "n_stimuli": 28535,
+    "shards": 24
+  },
+  "status": "completed",
+  "payload_sha256": "398c69f7e3074a4c0ea781e2174ea8c85d7a1cb9c9d21292019aea2f972b2df4",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX, Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T08:45:51+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_copco.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b7445ca1521cf54b9293f4a0642b9c98d7a58e0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_copco.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "59ff7a19ff6ee52d54672dfb12e75d3fe8f0e0a360850fb845180c6dd8a9548c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T11:42:23+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_dundee.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..d285ee5afc7c94f22751c39c30ba28bcf34101db
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_dundee.json
@@ -0,0 +1,70 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words",
+    "corpus_sha256": "55e4d5b6adb8f258a2d2e10aa3b9c8ba3cc43942bba3df2f1fe2ec337793c263",
+    "n_stimuli": 20,
+    "shards": 20
+  },
+  "status": "completed",
+  "payload_sha256": "7f2b073f518b1448d6e0a1a75228b8deb323655bc4e2377f4304a87d9b912ab7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S, Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T11:21:29+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..eecc9007f05e10a545cb3dfe9c90070084f37d35
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c18818e51fd0d16deaf10a20e2ea94cd9e6c1705d2a69abd6287892d0ebc4558",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:58:12+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_geco.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..d4dc220e77299a4d7dfa52bcfd2c54d2f9ee3585
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_geco.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "38fb55ffc3674cd33b6e79c6d066fb9e1097ad2586529016578f1e9876d2f6bd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T22:19:46+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_de.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..089ba88b5ec5d3bd6333baf14a32e06661fd7618
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_de.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a2c72efc2e5e0039fcb70081f245d53e125b94da2b4b2befd52061e631143dd1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:04:49+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ee.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..578d181f09ccc0cdf8084a26cafd1e661d1c61c3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7c7e059657fec75a81a5ee0494f3eb7c7da0e0c9eabd70db81623214ac5d440f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:50:55+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_fi.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..39ebd440c7e774ae3ec5879cde47c0e4c6742ab7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "9e49e8797872a17d7608538344d69fe7fddfb118bd800581858b61ee14541022",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:23:39+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_gr.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e46f96f76181d103dd6abb4f3d53213906b69c8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_gr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7a863024614c7bdd15d18739b8558f7dd0aff14e619fe1a1503122712fef4aed",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:30:53+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_it.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..65cc15e4500a9ccee3dbe5d3457b703b31178efe
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "00a92285e31405eff3dda5faafc4d541dfbef587ad9860cdc7e3daed3ccbf5bc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:10:46+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ko.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..cc068a85d68134a66fc0119faa295dfbf2cc0f6d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ko.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "70f03a970c6348e2751e32b47545e2ed0d7f52a802fe12b5b50c79a89f25552c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:16:25+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..35049fedf9af7875c3066ff206038ab0d36e6c43
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ba.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5b7436c62c1237d1ee5ce701a071e700a8e8b1a29e31428a3c6572e20c562171",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:03:51+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..ebfe448c1584078c038e157323206d23d99996c5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_bp.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7c0ae01c8bd13ae43031d8634e33a7ad4ae0383992f7e146aa0d4015faa2bd72",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:05:04+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..23a472da9774cd9f4cec945ba0cb60e34eaf9b63
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_po.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c7b24cd0aa931cc264bb3a162f70ab7d19dc1f257cfc21278e47f801ec356185",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:07:47+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..a058d8446c5d63234a3890c2a5284f0d2a263fab
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_zu.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c2840066403942620c38d38d9a1067f6cc794c04b1dede6e4a7b01d3a4c07c56",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:10:43+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..483b3cfd1fde060f650ab6e05cf8003c06c62ecd
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b2341fa4438343631b36680f0cd1ea952b7055365812e1a4b42d04dcda08f48d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:04:28+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e8094e8e8cb66e51c94ce9a0336c8667e25c358
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "949544ab193f38763167a820541815c58133cd2bf1725fa956e9358fc726aee2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:05:21+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..4eb67b2547976daa10024d3e10fbd191d453b652
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ic.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4a851e19d43e8741a6c211737336d6869304cae5c53c97f9a42b2a443a2fb470",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:52:20+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..6487d9375dcaacda4c18c92a5034196f4c7a5f32
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6d186f8b21a7c183663d27ea6792776e264b109d43cb785921d63d43dd0a1ff5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:29:06+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc434b840bebf5894b3ec4eba1fde2057019159f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7fb5098c320a5a00a45f4f403acb66ae55e5e269b0464a6aca66337689a5f000",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:50:18+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_tr.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..70e5405734b26d3d9496762d48053ece4dbfc2fd
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "bbac667b8aafa54a0b0c17603ad1468b1a113e6d4cb40be68404882dbcdae2ba",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:24:59+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_natural_stories.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..c6223c43935c583790bfd8ebb564b47f48b58e19
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_natural_stories.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5badd3ba84498320d968fbd97fec1320fc13623e138f3e3c6ccebafc2de9710e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T01:17:50+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_onestop.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..63a4fa6b785607bf7bd293fd9222e7d7a0eb20ff
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_onestop.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7d0e169d1766d157e9c2a8df56b970991a2f89af6c168edfa922c843612bc77a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T10:57:16+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_potec.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f411486f70e4fca14f00dafe7604a715bf6ac1f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_potec.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0b3bdaa6bd13dddefd889996c89100b37b65107290cc5a4062d919c2c65dcc09",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:26:27+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_provo.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf43661820f2eb84ccb4c25de7d1f419ee1726d5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_provo.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b5d31f517bc99862ecbc51c9f656095fe0e4f93ecd7760d39ef2d1eaa1762466",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:17:12+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_sbsat.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f05c1143a0755e9bcf2a75928b06ac3f180628c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6907713003e71ab62e6dd03d75d0896be33b74af2b8e087c3b19eb494e85835b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:31:10+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..074ea2194adf87a759d1deaa6b3923d1a6c0b17b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0e5020af33c1cf9f5bf151d82e40ed92d820edc00baf37dfedb2fc46bd08e127",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T22:42:09+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..40e67814623897fa7c74ef21f392b2bbde855e5b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "1c3cbcc614713b0d3457de26a7d5b547d0eb24e8dd5b92805e2e546ec9b74bae",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100 80GB PCIe",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:22:46+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_celer.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e5fe98a44619bdc5b46274da41df4366a4d24e7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_celer.json
@@ -0,0 +1,70 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words",
+    "corpus_sha256": "53cd9eb062ffb62069b53c2911e7bab7257ab54d1cb6a54b5cdd74304fbdf92d",
+    "n_stimuli": 28535,
+    "shards": 24
+  },
+  "status": "completed",
+  "payload_sha256": "f7485d9ce7aeb740d9a6c5ba26f2caef3777f60ad1359708cf31086b2d5bf294",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX, Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T08:39:13+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_copco.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..68856ed7894da635ae8b114d8973e049ff03f9f4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_copco.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "237fc8ea9a677287ea80daa492e8d165421f5b5fc69dc32c7e2d208a5a6d4f3d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T14:05:31+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_dundee.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..be899338ccb2eaab87790f024f76469c2b2d58b0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_dundee.json
@@ -0,0 +1,70 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words",
+    "corpus_sha256": "55e4d5b6adb8f258a2d2e10aa3b9c8ba3cc43942bba3df2f1fe2ec337793c263",
+    "n_stimuli": 20,
+    "shards": 20
+  },
+  "status": "completed",
+  "payload_sha256": "8bf66e6bc374828d7bae5e9784a1c5f7ae9f88ae59a2026435d6097307518fd7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S, Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T11:26:52+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..2210ca6526199332c4a23ac990769be4dc438028
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d8637294e741091b9e64f90cb800c7ac325ae4a68d74c6de43485ceab41b5af4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:52:33+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_geco.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..0cf1b1428129aaaaabf7910b98edfe7bc8fa7676
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_geco.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f916c9f02c5ed33c6af593fb9dfe2c91e1a5cdea5b3d958ab6d0967e0151f76e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T22:06:54+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_de.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..312e56b015533c8ab0ced4e0f829a0f48111d9ca
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_de.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6855f396402b8349409c233312c533d2245785dffb78e8b4991f2f29257ade3d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:04:00+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ee.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..6be03a5b10e2bac8f1b590095587b08ad3b67cd0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0b8708aa67c1b5818ccd4355ea32f31308e9111d79932c97e419c8a18a8858be",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:03:03+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_fi.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5d1871ae97a8207211784c9141cdf41b46abcbb
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6eac0f7b5f9f02ad9794ce8e738974833e2f6f0b000997b8747fef9286756b89",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:14:45+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_gr.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..24d0bb8888cc1b632d767b4e5f8ce4e1881b9d38
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_gr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7ad70cd7d76f5fd60909dcf66e4a0dba77a6f6c392b1ad1c58c76a05fd6d6f67",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:04:54+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_it.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b48982fd1d9fbfbf92350ff0e6c11eb72126999
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2c663bf81ef096ec0efd2a481ef2de5b9e6b1e08b52c56cf5ab4cb2b84b51a13",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:03:51+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ko.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e807fbecf8d77c63aa8b83f013e82286947569c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ko.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b6ba97ff44c7bb423c0fcaaa3526f225a9e3223e324105dfd0eeeda8007760b9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:47:57+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..2918116280a5a0435493d330186aa335c2f0ec84
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ba.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "9474798aebfe458ab7572a381e5b03f99d32065fbcd53e579d247251e2a9d3b4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:36:19+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..58bf54e43f584ac792520c3ddb6f80af35496f00
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_bp.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "050beb7290ce5a79c1c39d1431aef6d57932a9ee08f179c98478d56d67793332",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:07:41+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..af2205e009c564b62babc86e55610df85d451939
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_po.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d94eae803b2098b43c83d24ed1e871eaa90d266e8225449138b732e43e6860c8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:11:08+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..cc6cbd2e9bfb025d8d4ab3b5c8a69debc6963ec5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_zu.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e82ed6d3472d6f614af8b0f12a47e7ce339f9b88285e43f9b1eac8cdbcdb7eee",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:10:54+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c0c250fd81080b9e3286f81c81e81d94d67253f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d6b7f29aeeade6eda060989b8d64bbbaaf8e46df9d8f5e0821680579b112fc17",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:04:54+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..17d5af97038c97657b47c28d40ea2258226d5979
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f4e816930daa7b1ce0d7753a606c103e93fae0dfbb647e5a5fc1e7ea61b757a5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:04:17+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..85921d84360d768d87e9a5ee28062f2e55196bd1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ic.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "29e7868fa7dbaeb350e22e9add88c6e3d6528afb9b42849034f2f56abe67b467",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:56:12+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..cb1932c95af1b0017acc982f6451fce48f7e3e64
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a602bbb7326a70b9cbf5e0d1ae90b0f2924215823f094f7e43578b912756039c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:32:22+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..695aff923ec5a126622b5c7d696a39e70a544894
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8bf08702010a683e1cc1477bd4661722ff068449d820936b613cdd342ba5eccb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:51:39+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_tr.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..08a002e12700383cde531655c202295429cf795f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "82534c76709488d090489c9d3ed8473a31f5becda790a02dc9aeac2e30e25a16",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:24:34+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_natural_stories.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..1e7a6e444c738706747ff31c6f79212731c3c3c5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_natural_stories.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6933d1eefd3f388f2d2e247d44e4bf5159be58209a436e5ca34b4241ef2ccd24",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T01:30:45+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_onestop.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..549699b57dea68b96f0d8451b88c379178537fcd
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_onestop.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4c9a877c252edc8bfbf5109ba8194e1609c7688f76d486a609bf52eb1de98294",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T11:17:04+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_potec.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..574045d2f2b66c0b63faecd26f5a282082ca802d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_potec.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5cb808bb15991ee9cd64952c449618632d3eac2767705ccf73145e62778dfe1b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:24:56+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_provo.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..b647130f1a4c4562848ce352200bd32d460f0be4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_provo.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a9d2f6d5a6e6e4d33cbddb7bff7dcfc5259708499fb9de6d1ffcc28752f252da",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:31:08+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_sbsat.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..a7e09a424221e0897d18d86702ee683894df97f3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "57aa7e85ba5d76aa213152755646b3d763ee3f021d49e716cccf0c7d559a98e8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:39:10+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_story_sbsat.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..87a66358051aa50c21e4d757c7e76974a8eedffe
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_story_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0a35ff2b9bf0a121822130a93b4a60d810adcdca83884e4a2c2aba767908058b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T22:34:00+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0f00c4116a2c7fe48ab11522a567e6d44f77a9b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d226b8051012182e533366aab29dce32f4982e37a028f120dd9a8efc10747f78",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:47:58+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_celer.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..d49858fb671d20a16730293afd0ce08fb009b5df
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_celer.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "33eddd4837a02ac16c9005c759ef03f8ae0daac00933ff68379ff8c6aa405b07",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T13:24:58+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_dundee.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..21d9036be640ff87469e90fece9b7b0be2c2f425
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_dundee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c39629911b50d52c255cc4d201b69f39708db0d96e4e0c66b91984d69eba8c14",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T21:31:53+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_geco.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a8d1f6ca62dc4b83e2605a5bd49b97fff150bd8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_geco.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ab6c2f25c0cc7e4e40bc2856d332b3cee83c95249bed4bfc94248c0e7ed9339b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:11:59+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_de.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..e67739941664f1a837e616baafa315c1340f6755
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_de.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0f40473ffc701791dd56535216349e86fe43cf38a9168c8c1d22e05052158de7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:33:35+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ee.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ed719b986eaea80320ca1303c04bfc1ef66cf83
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "75332d8a5b66509684fa7588ca1e79f8d679f2268a10d5f93d0c83c108b6e755",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:27:54+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_fi.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..c219b3df1cdb9b2aba8d811074760b7ab6f1a3df
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "440078bd7ba073ed6c1af3d522cf903c45e0520a7dbc48b545b55081d6587fbf",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:59:28+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_gr.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..3bdf460fe14039ca09e67bf53f86a09037903922
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_gr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5736ff5f536d862c5e58049e35fb6c0428c43c6ed1d9e9c497b1655353ebf396",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:03:19+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_it.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..6eb61e488221d682a50187a806fed3e35ae49522
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "210020db5703ece27b8c9fb8a32072cb10e99e2519d36fea3f9641f31300c9aa",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:33:18+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ko.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..d78922f25e21da54236c1aac6b8f5f9d843aa6be
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ko.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5e52d774843e7f39ed15f37ea5a671f7f5a677234c84e9f28f9fdb5cc324425e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100 80GB PCIe",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:21:02+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_tr.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ae060f2d59c4ac23e9d321ca01507e2eb2c5220
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8ac770a110dcbc6defdbe8e01ab200fe2a25c3cce832e9e028324445e939cd19",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:41:38+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_natural_stories.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..d471d5349e591d5a3cd0421ca08cf5d23c6063a3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_natural_stories.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "9b3493964f5855c33cf5878ddff5eeb2597d75a7f23642225caff76bc681ada5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:55:26+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_potec.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..213435a2a2264384c7a03537f00032ffad23335a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_potec.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8b790ffe5e839b1949b80b2d5902323e0810371285c95ad6d21d3760ee1eb35f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:44:01+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_provo.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..08fbe7526939ff0da885afaad22a97e55a6ffc60
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_provo.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a0e163618482af99654c6d137e3b514bd559bc033c382930f100c6eaa3649b27",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:39:35+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_sbsat.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..312ae520ed735f872fd83caadbbf159d15da5866
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0797a537762b8e1fcd67992d1da3a3ee649586e6d86436b128c5763f98dacc97",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:58:30+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..5b6cc0cc7026f8d27858dd0dcf656d1dc3881f0e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c93255dde0ac9520217001c8a2785a5ef87c0c9382c11e37545607e2410879b9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:53:28+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_celer.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..2821fd7e86d2eb6c74d858a46169b6e2835f2f5c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_celer.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "daa293087cc20d70bc6384f55848b0ff8c4f638116c6bc114e53492cba5ee70c",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T13:20:03+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_dundee.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..2de53349a8e092a00162b6e8a784a650b2fb25fb
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_dundee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e9ddf0ac275cf79bde25218d5883c3209b88db32c29046270b8dc7bc99412269",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T21:49:21+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_geco.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..08e19d662cc7480e30aef3839e7c7110c65361e6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_geco.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6e1e22986c8248c8a5e029ef87d1d88c5bf8b53d3034705bc10deff8313521b3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:11:08+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_de.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..e910991a6a1514d0f46bc58fc73bf3248b534059
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_de.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "3801a1f923bcece7d2b9852e039b3e6e8d84d44add6613ab2745f47da26c8d98",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:34:20+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ee.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..948957de9fc0dadc4fd459e27a607149a5a6d117
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ca3951e9c6946a5ca7d475364ffbeb42b66e2e939ee46d59b982adf457d94d8e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:15:59+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_fi.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..6dd77404482f1ee8eaffd72c240b11ccf40d3c84
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "87efd52769de2f7c03887c3e7cc2b287e0f69e0c94fd72eb89745f86c90191c8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:36:29+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_gr.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..0d611fb57c72adcd35a827a7e340a06a75ddf30d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_gr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2f82b23d7be2aa9185f533e2c8d7e20e362cb34c2b20f3be9b1ae4219c1576f3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:43:30+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_it.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ac2262d5e28fa23dfd738b8eafa399981ee0959
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e0fad7fa9c1489ae0366c33928b7f23caec1729a1dd89bd124730d8497136d51",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:57:39+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ko.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..1c1434384f534fe5721c352fae9a037c623b2ee1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ko.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "eb5112ab7702dfafa40d854aba466925790759324236d179a8b2163b625d4e77",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:56:01+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_tr.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..fb3468ed55205199bbc0fc0f2cc0037311852e93
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ab589a30344e7d5652b6b88f5c5f09e80cbd6fc1c4e0a0327ced63911f9078ba",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:41:25+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_natural_stories.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..220b03a37dafedfcccdbb22f6d45ecaa50015782
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_natural_stories.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "434a65a35a50aa94f6bd08f9cf9670b3258f27ddfec69cc1b81a7e68599fc9b2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:56:51+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_potec.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8ac4f47cec48771c238008c0487ca9f3c114f96
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_potec.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "50ce1edce10790f50055ecc491fcec6b41ec5201e6c702c4f1af14b4fba2d3ad",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:40:22+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_provo.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..693a0ffd30eceb68dfcfe93cd800b89b1641dd50
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_provo.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "9ae0701ef21f9044ac3c84bafa68a80f238e5ab15a7e502e3b3d52ae7afc7266",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:45:36+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_sbsat.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..1732e384d9545ccfdaa4f9e1d9bf215da5116da3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ee68e845949ced21fb3d0fe562681f655f2572b269cc6913b01924e9ac1dbe28",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:57:48+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_story_sbsat.json b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..7798a74ebc3b684f388361e2ecfe0ead3b8df6ca
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_story_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8a94d14d62d7f32fe891cc423a7ae0644fb7bc2e8e387424345aa3316522fbcc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:56:48+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mmc_unigram_surprisal_federmeier2007_n400.json b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..e223a0156f4c8c99615ce46e7059dd17ec01100e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_federmeier2007_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "09b4627a9c1eb5163789825fc88e563aeb93e89093dde7d9634e1dcb9e7d5e2f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T10:28:26+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_de.json b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..98bb3df05f7567c7403234a772ecede211813680
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_de.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "62d94be4b7269a3fca0e6e1f993b530668fa0e78031a56c51972a4ac47d3cf23",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T08:12:46+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_fi.json b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd69562e78c17e4c9da81102a39455dcef44a726
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "04a6c9cdfaa77456e4eb030c057334cf20d174ca186ede7df3dbab26110f8d41",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T19:11:30+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_it.json b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..45eb63ec266b26821dff128c08c8149ccda82b00
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "3117b2d8641335032d86c166745c4fccb1bf6aa6e0ce438ad40c08d98e53ccd8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T08:40:02+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_tr.json b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..9faadbf962260c96b88a651fd4574a8d97a8c454
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "509b55fe0cb747c8e68844e8f11f290627cbee2103731dfbc31fcb6a6988d1e4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T22:33:59+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mmc_unigram_surprisal_natural_stories.json b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf180bcd748813f72cb6c3a640dbdc43a7d7ae1c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_natural_stories.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "955411cac2f344828ed4aaf664e82bb7bc41d33818a0c81159119cfc2faefa1e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T03:53:43+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mmc_unigram_surprisal_potec.json b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..087c0ab02be51bf12aa1a6139ad5084c861e5097
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_potec.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "32a8335bc17cbcfbe1b69717c05f9ee8e91b7ad3ea56ca7855647d31ff84ff1f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T11:28:40+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mmc_unigram_surprisal_provo.json b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..ecdab9f7ebe5022fd27a265697b994f6fa8d8ac4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_provo.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "3bdc86b2e27f357fb167b62142b65dfecf02cc659c5dadb80eb1f22e185c3512",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T03:19:59+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mmc_unigram_surprisal_sbsat.json b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..762763f2148cdebfd02a4c332cb1844235aefbc8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "3be1fb4499817705dda625a5d277286dcc44b2aa38ec93854188ce3f33890a26",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T19:44:10+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mmc_unigram_surprisal_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..42674cd2a1fc42899d4db40daa34117ef2f5ede6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f6bf345cafadf1a17bc794d8c04b6e4b114a88190bc544dd1e295f5b06b1bd5a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T11:18:43+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_mmc_unigram_surprisal_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5f6e335822717aa6642dead84238e4446c2acb5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_mmc_unigram_surprisal_wlotko_federmeier2012_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "dac766254eeb05ba5fdb5ba317dbd2b66616b54912ffb88246e2d756049ecb49",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T02:46:15+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_celer.json b/manifests/ai-forever_mGPT_next_word_entropy_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f5bd11ff52aec215cc160ad7b3f7dc66e95b102
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_celer.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "179f6cd906d677cea5bc6f79fe696ae7506e42e7b39bec0bfe42ff75a7b6801a",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:02:36+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_copco.json b/manifests/ai-forever_mGPT_next_word_entropy_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..c07b8e09b4f7a5c18bfccee1f68f67b4097339e5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_copco.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "33d0f562bf91569a483f2a88011ae41c08939639bf393082af201ca33c18de5b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:05:27+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_dundee.json b/manifests/ai-forever_mGPT_next_word_entropy_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..9de1ac43c2ae2492171df80e3a8360759e38b207
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_dundee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "35a0e1b22c421df623be6adfc27263f42abd0a0ba3fa12702ae757f6a8ef07b1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:11:50+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_federmeier2007_n400.json b/manifests/ai-forever_mGPT_next_word_entropy_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d55c2ee54e6afc8af15b650d537d2c5bd26fe39
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_federmeier2007_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "609a9a75c05e839c5e3c5854124494fb7b80dc6d484f871f5a66f43a3d51b9c9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:17+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_geco.json b/manifests/ai-forever_mGPT_next_word_entropy_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..a56f067fea85f7beab691b7f5dd55ea307f49cc5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_geco.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "84f4d2214828f4f9e0682ac07d8ceb7cf63a4f114103709eb7c2a0168dce994a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:35:33+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_de.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..03aa306e47b70186d29314fb109ad21926b41bad
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_de.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "26f9026255fda36c8784f948c9d42ea39c7407a5aa7fa79b5cfd6e66942c9902",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:43:32+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_ee.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..b741cad05083d91c1ab7a800cd1a9f297ef79f84
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_ee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "46f93edb833c647b6320054eab7aa37ac7bdc74c5d3ee7e7e15caf3514e4b8e4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:12+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_fi.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..91b8df9ca37ecb8c60582df5ed87a97e3d5e1b48
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "38511f7759cd6197e16babd91569ee000bcc4b877f16ebe79d743c5c356b4e8c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:10+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_gr.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..571eb82552720a93cfd6d3aff088841554f624f9
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_gr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "53e71151c26e2a7a1987a9d921214e38b645bbaed1f79ec01efa9ec031eb8a4e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:34+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_it.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..6253b9e7c03c8cba8333293c3a477ce2f744791c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "cc66a39b7567be841dbc73700cd67b1ec95078b045d3db38c4fb7c00df643456",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:34+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_ko.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..1730230d004b8840ba5db9a956317e9c013f2ca4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_ko.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0f98a77fd737329303279e31ce9249665c338923ed14629a99af123cb41f6a98",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:45:13+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..4bfa0bfbc2a3b67e2a0b54c45001dc6f8c8bfd24
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ba.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "af1178bb382bd1d6e2b1566bf2775762a98247481d9dae751559bcc9b9ac7fab",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:12+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7e9f6bda42925f6651f79d1ca6625fa38bed034
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_bp.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a6fd98fde7b5efc476f1336b4c8492f522ff266bb93d7e8d536beef94971e2b6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:39+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ch_s.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ch_s.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6ff35981f64c91e181339f283ad3cb0a6a38807
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ch_s.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9f27685ea775f0843723c5aa8c5b226b7d3420ccb213afd1eb8198439dcdb6a3",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4081b1818d57f69b4674ecd9e1d794a16f9b6a1ec94d5c1191a18823d88db379",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:46:13+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ch_t.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ch_t.json
new file mode 100644
index 0000000000000000000000000000000000000000..366ef58a8bd4fd345dc9129ec93a6e5a9639a8f7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ch_t.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "866884887c127db9b93a7cf7e4deb154d4c2cba960aa58ede0d05493044e3877",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "1fe448b0cd9d97f7136706d60e6737b68766370e7e9f6f4c90ba97c6046914e2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:46:22+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..ed1de45891a1b14e11ee7ccb8b1f310dcd6d2696
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ge_po.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "dc07cd7670ba1909975f3822cf82889383079390e131b0b84521dde9fedd1a77",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:55+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..47ffe0d33d0d635b9d6b5a6d7a1476531e1d6744
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ge_zu.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ebc3f39e709d4946ff61c23528922f60829b579500f51f1cc9a69f1247bff705",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:55+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..04fe6fbe9dddb14877eded6fcc15cd3b8de17a9a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4085e4a7f204b71b22572ef114cb31c98eb16b9cc3383e400e6b9be11e6625d5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:12+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..3153194ba66436e276e209bb27a5448c4ee167fb
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b8e8392bcabcba26dc037f75553e237245efe8ef087525a5f4952177de28ef8f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:12+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..98570c5db6c6ddfff49f38c42d3a64c2abfc05eb
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_ic.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "469b48e8074849fede73e274b02251aa1f8279b16f2a6e38585bd8f66ed323e5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:32+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..546e60079d341ff6467e831de7fcf0a73efcd81b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_l1_w2_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "904aa4782930b6d48c6c691811a9db66c1677a6f27d6d2243d0a82649788996b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:50+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_l2_w2.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..87eb65a9bbf3b97af03b6d7adb930f24e7f52973
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_l2_w2.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4c145af9d62106224b1c37a18651b31f1803d85e708f12a85f1f714b698842c0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:58+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_meco_tr.json b/manifests/ai-forever_mGPT_next_word_entropy_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4e9e58ded36f2eff724e3e0f83a30f532438ad7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "eb224d24cb571ce1343ca6014ffb7e52c0783f3efd3726a8027f2b68792b0d1a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:08+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_natural_stories.json b/manifests/ai-forever_mGPT_next_word_entropy_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..ede4fe7fc6615f6bc2871f3fb7be9c01a0950c3c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_natural_stories.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7301e4ce1781796ea873aeb6982fe29548d623296de1089e9bac84d320d0d947",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:05+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_natural_stories_maze.json b/manifests/ai-forever_mGPT_next_word_entropy_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5ae33b52c73c2ecc7122a868560ae0729003e43
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_natural_stories_maze.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8369a8691eb75d9f09c5d7c3209b82e0a3b5731154feb2c3d3205e4c83f008d4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:00+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_de.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac81bb5689700bebc0ee6417081792e8939327e3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_de.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "31977d7531dc05e2b727439826abf01e16b2e0453740e9298d8cbf272491784a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:42:56+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_ee.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..b36e7ce44aed5f5587e4aab95161c134e476e863
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_ee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6b9d881ecb4322edd9623d24e8a813024ad0005b97bf1cc8e3b9345b8f720383",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:12+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_fi.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..30f650056a0816704ab609e9b5381a09d30966b9
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e51a3d2bee0a8ce243515283f7d5b1760fb780102e616230bf6871c1dc30524d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:10+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_gr.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..07c7c1f7fdb558e115243d12431b327175338041
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_gr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d5b24804af02af33a146080cfb30351266e7593af0e60ae14b28cb6804affc83",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:34+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_it.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..473958bde1f4cbaaca9bb061c4bea6b4f41ae6a2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5b4fa9a4a5adbd2e89bbdd7ccecb88deed97ddfede60e8a82344708bd566a748",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:34+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_ko.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..a4d67940359e8c86798441dc1bbd9ac3680ff09c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_ko.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b80a481a88436a01ecd027b803cb67907d42b5ff61ca4407d662eec911f34bae",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:45:11+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_tr.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..1673d22972553214d3edae686f3a81ff8b14fcca
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "85b2f94158c3123c849b370b1cb2fd3ba1722e2b76dc001b4d7fbd6c1985a5b6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:59+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_de.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..1df6860ca91b204f715cf00c03ab2e5a40b06641
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_de.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "8a91f15c9c999ca3d2b81d5f0f9033275054cffbf4dac78e99339230ee70057d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d7552ee4846ad170f376fc16cb64df9d804f9bf7b9516ff1d1672b8c42d34a79",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:42:15+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_ee.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..6633414f3c7778d5221384fd17ac2fc911ea05d6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_ee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a3bd7f10daded68cea3728334385d0638ded96ba61df00cc85eb4c98848aaaf5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "313b6c815a1832b354175928d72feb851186a916171bf438299c75a689161904",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:11+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_fi.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec45d6668fe4b30c1959bf416b9b5aaab3d4e0e6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4ea11b1a9e4318635b18087b3844bd3cfa3891f932742f2a3e93c0887c79b57a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7d539317f1355030c600b395ee2ac316409c7a0ab4906e11aa8d425949f56c45",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:11+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_gr.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5bb8b8f315491bd8c8611dad56b9e3883d08607
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_gr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3fea08cc811f5cd24cbe536ebc4360c1a6dabad4819c7515a4cd6b68ca8155d2",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "86349f5f560a6ada746d83cb6de6e12ecc66e65c8af4c2ffdbdc5a0a9c0b66e0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:35+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_it.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..2dc43aec75abc6f9c94930608c64c727d82df9d1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "6da2884eee23d85230abb29366689af0289e5e99fc41fecba935ccd3ed55b148",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "55c313b5f1b45ddbed90caef1e248ef675151111808cb40827a4d9d739530129",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:47+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_ko.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..fa5ee7af5eca6c1ab571f5ac1e621984235acbe5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_ko.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "8a309737fd1a038f8efc74d3e3f6aa24d3415f1e8e7aca2ee3d39feccef6b720",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "312f1cba67d57a094c8374090bb850604a37a0d3da46c3a0e1f5e0ce2ad55532",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:45:01+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_tr.json b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..9263cc3e982539bffd37a9637be768b381624ee8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_nobos_sent_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bc412857c22dd42bc7497421e05d56dccb15e61a21e1eb9925554ec022807feb",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2a894a103b318b76016c8a3a8957c62503af29673ec0d13d067513baad8b5fd4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:39+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_onestop.json b/manifests/ai-forever_mGPT_next_word_entropy_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..79c31fb7b0a81b0b8b514099574821307be8270b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_onestop.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "665a06578c1ecdf5b911b2fd9ad6ebeca2b34edb51f84f343a92afb7c67086a4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:56+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_potec.json b/manifests/ai-forever_mGPT_next_word_entropy_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..3312c077b202e93f3433ccf4730d69bda3e9582c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_potec.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f301dce524a0f1c9f32cea2e4e297dd2a31eaf7d6dc9c5dde081111ec1a3df32",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:37+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_provo.json b/manifests/ai-forever_mGPT_next_word_entropy_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..32ea324a0c268fcfb64d80133d7641d4463706f4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_provo.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "030625de58c3a489948fd40fc1f8b4a148f3c547b977e2480c00fbed7534a65b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:00+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_sbsat.json b/manifests/ai-forever_mGPT_next_word_entropy_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..83b3d6737c21f719e09589645bb1435fc4a132ad
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "9ab07a7756f2ecd440f7306c2c4ec1fbd656c5dd50b123d29f5a311f2192cf10",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:46:55+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_story_sbsat.json b/manifests/ai-forever_mGPT_next_word_entropy_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd9860bb845e14f5bde4649c6c5617e4edeae638
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_story_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6c23f115e873f300f72beeec371bc9255acd82bec9bf2cb63156e8c2d8f120a2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:46:36+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_next_word_entropy_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..18bf7da02e1b64227e828bcf46b55cdaf604a9b9
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "62d4c43ed2903c2e8d60a8a72ee11949a00fb3d560cc269e37336b9cfc1dbdc3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:18+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_szewczyk_n400.json b/manifests/ai-forever_mGPT_next_word_entropy_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..98cb4f126da3528a99beecd0d5798d737d1addaa
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_szewczyk_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "89a8746ba6027621bcaed51602168bd2b7bd941e13e8eb2e411e210c58302ca8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:39+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_next_word_entropy_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..9bbdf5e7fbb97d2ea4d225f46e07039c0b471e5b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_wlotko_federmeier2012_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "97023f90b1fd8fb61b5d163adc9d6cf04cfbbe23498a221e97a07f981b92ae4e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:14+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_entropy_zuco1.json b/manifests/ai-forever_mGPT_next_word_entropy_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..6336a47afe1896e5081aeffc878732c69c8411c4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_entropy_zuco1.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "176185c437d6d60264daf2f0a85bee75a9a424336f6f0a057a5d2b6c616eb0d6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:38+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_celer.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..78f78abd1d1f55c12a87606f06dc2610ae140ba4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_celer.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f62f11bf9a849b6f4625a5251cb83ad4d74ccb89271e2f70db72d7e1b64fea6b",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:11:21+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_copco.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..35e0ff7164291b421847d00dfff09949719ccf65
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_copco.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0665bb55ce474072c675009353bd4efd75b5bc05c8ec53d4dd0bef0524bd3a3a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:07:28+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_dundee.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..e37bc1cdff90cb2e25f4156bd28ab09338176bff
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_dundee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "050bda0112aeae2c76fe79c57ba112ae286a859565f9829c987ca80cb04822d6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:12:45+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_federmeier2007_n400.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..3af027baefafa3e7644005c43a62e8d3e44cc0c1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_federmeier2007_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "297967d8dddae700aba3b45f0e2c6fea8c39eeceb007ab79bd5b70bfdb5c70f2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:07+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_geco.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d489c7ca9138e45dc7ddfc788fdbf72f6c21e6f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_geco.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "3813f6b725e812d6f7fa9d3f6d9f100d06b0684f2aa6c2b7f5f36a17f7453518",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:39:03+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_de.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d1c73f42b3768ecc7a5a75591eeb991d603bed7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_de.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "3ec1c854c7add600cd9044c8a830600b3d9487b401642994c921003c9539c0be",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:44:08+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_ee.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..dfa932c72fc10469316005c8a89fe62e62d266ce
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_ee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "51571fa4d44c6ba217c244fc4a9cd309cdca818fd6539e25abf8a5176ccde5ee",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:45+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_fi.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..416b545301242008959a714764b6cad3b4303bb6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4564c2275907ffadb7b99fcbbd6e6d31b095790f805d0f3ab96e3359688a0233",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:12+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_gr.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..65ad5bfbd6e0ae88ae0d96fe5a11c050f0d5fde8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_gr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "981d494d3ce30257178d4a27d568d9f15eabf9a9e49918e0234baa0b0a64b149",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:13+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_it.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..f62006d6070ebb172ecbd2749976bd00ae7ba81f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8760dd82c86f119af947a026a450d9db635e1defdabecc1f1b7a376b97c6dca1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:36+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_ko.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..8276f618bde403d535ffd7c5f1fc468940c359f2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_ko.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "944112f1843ac22fe4f9275cf84ccaf669f9f1afa012e33ba227a3d1d724ea65",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:45:35+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..15e4c487cee5ef268bf08dfa8b10cd798c82e5f3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ba.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a5da9faf42287e87aee08f14271bafa8dfa03d02add7b8fad2191cce7f8b9661",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:28+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c4963b3bc1c3cd365178c33523e4c5a03293600
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_bp.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a26c28135c394dcac941889644c9f2bcc554de94d66d934e40dd50fcb736e556",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:24+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ch_s.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ch_s.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f51099fecabd858eb460be83a2a037a2c12b26a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ch_s.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9f27685ea775f0843723c5aa8c5b226b7d3420ccb213afd1eb8198439dcdb6a3",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "347a406182df2fb88e0edc199419fad17bfeadda2007d455f51a1d23b2dae7d0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:46:17+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ch_t.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ch_t.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ac7b05b7b2be9bbb4bd1f32e8db97fe92f2acae
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ch_t.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "866884887c127db9b93a7cf7e4deb154d4c2cba960aa58ede0d05493044e3877",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "660add0adf45764fa1b7b630fca5335147d70de4e6f47244ea91b91a39e993e8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:46:23+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf75cba4e92a340206ce1e296815be5496ecbad8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ge_po.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "096ecc6157a0211cd2eb25b85257c448a3195852ef5ce26e7f9652c213dc5d97",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:11+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..a6a3c503ff85dba03fa9c472fd722b1c1482e79d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ge_zu.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "19c2afaddcef93c35c627f873cf5f1a39d2538df9e2ede214ccf8bbe1016d2cc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:24+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..81fc3a44398f6e4bd7b3b51fce24356988e5774c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2dd3066aa323a399969d6c6d436feccb032b3bca4afb6907bae794a066041662",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:27+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c92ceb37cdf47a93c14909ed21865cfcd17e9ad
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8358473c8b59531306976c99d30abf1ed9ce1042bc71dd672c9fc0a39aab0a43",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:27+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..4123df65059a0731cd58397f74c5ff1c623ccb10
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_ic.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "988497d7b31e22104bad1c76b9151c9bf53775227d95a1625d16d05cee41fcbd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:25+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..2131ee7f20c3dff970025160c1562e3e89e25040
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l1_w2_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ddbe0a34c7aeb12aad0fa84dc5d0b319d858c5ff8addabfeb9543acb71ea6c70",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:12+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l2_w2.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..6bdf6378d5886930b5da1922397b15f142e4af15
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_l2_w2.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "df09658a58cf58878777351f99d39fdae1f380699f72694d0956b9b3cc5d3f54",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:04+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_tr.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..98c3651d88f4a0f3553459e55cc1ddc9cb1b2976
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "87e27d19511adb942efad14a7d02ff55078eae997dfe0a53f33f38fc52ab6f3e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:57+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_natural_stories.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..8ae1ec742eaf8c52f86729de69d7245b803f8aa4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_natural_stories.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "dac8c349a541045d9ce61d7141e435bb85b56b2a6ede0b8eaf68a8cb44bf1cea",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:08+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_natural_stories_maze.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..149bb1da1dc7df1210f12dfe9347f923aaa3c65c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_natural_stories_maze.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "53211d3afdde04db783fe1fc051fedb8042cb5826140ac17e4d3a1681d0ca3b5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:40+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_onestop.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4da543267d444cbffa5f44c51f952b99682f9bb
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_onestop.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7f1b036e32d641717034710e0d09ed7f559be925580216a8e860c41f1f5d898a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:22:11+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_potec.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..f35bda00787f9be545562841c97c1d345fa476e3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_potec.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4bb46872f22cb1fccd13b31bf6621cdb747b03e6b04dc4f7b42d987a14190bab",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:26+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_provo.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..70c2ae0d126b2917b40922017bd9a4a4c925f752
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_provo.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b1afe63d7c419817ff19c4ece2d40487a057b86de779a437628260653a858053",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:46+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_sbsat.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f2105e2a9b41adcda939727dd28527b95ce5dee
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ae864bf25e2fbfc28278548aca5c524b1ff0511917a1bc94a71cdc16689471aa",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:46:45+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_story_sbsat.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..f254ea458bc2283acc12029085aa27de56f8497d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_story_sbsat.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8aebed85cb6ecf1525ae19060271a6e606876d9b1dfa742b88691a4d9a640ed0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:46:50+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..c8882fbedea3de220aafbc7e53073d50d67c23aa
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "978a090a990f2197bca3a0ffef2b6afdb7ffd36b69b8b91a7beb3ac1cb98fc99",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:22+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_szewczyk_n400.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf008ae3953ba362f5f1bfd65e499aca2eb7d975
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_szewczyk_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7052e279560f07bac89c8374c8520e79950c3e1fbf258325cc55cc1699b9576a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:15+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..26adab376cebe397073f674307da2ba419bfc86b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "95d61fe12d582359f8cefac3f1fdd095f1f88229d330d4fc9294836f879a1dfa",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:59+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_zuco1.json b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c5ad25b011fc422ddcc562a075cbde6f32bb06a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_next_word_renyi_entropy_a0.5_zuco1.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2f3f14f0e3e23f10a8a095a8e96273853a051da52dded0d5edf06fde76085450",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:41+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_celer.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..9fe335ebff39ab90b53179962dfd2b65f6e62c5f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_celer.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "80cc6f7b16307917a7770530be9956102373334bddc3664cfbb5a8a767da1b38",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:33:29+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_celer.pkl",
+    "source_sha256": "c26e750a44fd28563bce0f935f1c799a7dc42b1268ae2f34192f8b6de682cb0e",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_copco.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c4fce00df906ebe5e8be039ef3a2e4cf1e978b4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_copco.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2aa73134b04e7f9f6761bbc6af6a8ae5e16f1ff01874ad4dc75da14274094893",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:28:41+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_copco.pkl",
+    "source_sha256": "0c6484856ef76ab6624f983e86644729636352f51bc896b4a902838182c8154c",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_dundee.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..8bfdc8fed800e2b62bfc120cbb8a8e110465012e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_dundee.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "efeb16291ffc008676fb2be5a7c5537c65cf4c0ebc26f9c905f4006124e46e75",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:35:39+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_dundee.pkl",
+    "source_sha256": "832ab7a7047c77ce5193b676c2d235c27a8a5c4695499223facd337cb1ba9ef9",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_geco.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..2026a0240055edd71b9aa38db42858c75fe7f633
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_geco.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5eedf516c4e93078192cd868f1a4087bef53f5c3a3a2ed93fc8ed920fe0bc40c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:46:51+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_geco.pkl",
+    "source_sha256": "22b24b3f93dea7eea54420c6d6e85ebaeeeb247153e87f28b40f744cd77287b6",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_de.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..c25e1f76e207fd5cdf05f1af1a5a58a02d38f727
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_de.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "be8d037d3a6ad1a831d778a4ef9d50ec075a28a09c230c135b93f2b7fe048943",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:57:19+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_meco_de.pkl",
+    "source_sha256": "fb3781b289731e0ed43642be133f15727fe10379220c756b461bfdb4f4b3119a",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_ee.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c8105154528403b96fa9be44afb31c5be9e2cb6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_ee.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "13e295a62df1d0213bda89137af833dd59355d7097b019cbe1debe06d17d8f5d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:59:29+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_meco_ee.pkl",
+    "source_sha256": "d18d53fa2cf026f9dc7b0827c590e2e8f362cfc29b7b1d44ab8601b25b7d8678",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_fi.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..02f56772dd3449587796d345375bd9557665fdd2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_fi.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bb6810738a9b39cb1a1335c60fcb92d3e1e7f5eec4d6fdb6326b6e034c7dc90a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:01:40+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_meco_fi.pkl",
+    "source_sha256": "841784869073c6bc28eb919f46206d3f4f63daa5618363fe96eeb489f70dd4be",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_gr.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a247f50530b4dccbfcd92c25c80e4545559aa48
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_gr.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d49476b2b614fc8c06f17021eba664f438bbe5b20eb115de5492ac20c9369b35",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:03:37+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_meco_gr.pkl",
+    "source_sha256": "2ac8d55d4c889442acee449931c3aa7ec39b068d37dfe2b618a8f2ac25c74bac",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_it.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..b8855174ad7142f865aed012b81bcbd1fc60242e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_it.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d1ec9826c4736d86987a884319d9e8fff8ef58353f156c3bee5b9e5fdbd5f643",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:05:26+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_meco_it.pkl",
+    "source_sha256": "08c2d937b40281818d6cba824edfb2b0439a1613fba4080cdae64a508722f4ea",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_ko.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..99a16eaa44907e20b49eedaea143e2f8b887bbdb
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_ko.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fbbb51a6da5082b4eb55a858e72df9c920d382c0f607b6a29e434180b837bb33",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:07:20+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_meco_ko.pkl",
+    "source_sha256": "16091d1a0e32013a90470a7e17acb7f7925ea42d78ee3a591ad878a7b401e02b",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_tr.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..f79c6f6a66c04799cb3c2f8cfa1b6f45a21227ae
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_meco_tr.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "309d819faa5e4e0392d08b06c46d23511fa08bc77543741d3dc0097bd0c4efe6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:11:00+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_meco_tr.pkl",
+    "source_sha256": "c875178a11dc511eace562c6ddbdcf183b778ee6ff0f056ed4c10e6d50675918",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_natural_stories.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c7b08f31e6bb642c30d9a78888ab962439b78b2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_natural_stories.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c70142e8623d1151af4e191ca4a4a71e9cea7abb0d3aacb99cd6742e32b09ee8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:13:23+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_natural_stories.pkl",
+    "source_sha256": "143f17ac8700c71e8793dbdeb2079053db0b3b3334b3566f0215b730c3f223a9",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_onestop.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4e80c56b244f33e8bfe0d20ced63dddb5e5590e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_onestop.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e2ab7a2da9ccd9cc709af882f047146e0ed2c004df48753eb07539967c369766",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:19:33+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_onestop.pkl",
+    "source_sha256": "041fd7220e5750a3da00564de0895e017e2bd1803169bf15dee33fe683bbfc4c",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_potec.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f1b2e7c0633530eecf654df4a02bb4f7d323cd0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_potec.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cbd8884d0f0d77ffc76db93395c7a0942bcbefc829e02ecae0c2a7914abf32f7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:20:59+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_potec.pkl",
+    "source_sha256": "69015cc48e26d81676a78b631f1a7d8f4b933de537583077c97fd485c8fd3bf5",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_provo.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a940e9a7a7fcf1d05acfbcd1156601b2dc29b9d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_provo.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "58356cdb84b2ab9cea2c8fce8f34735f6740ae737038887adfd8703c1d6a7742",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:21:47+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_provo.pkl",
+    "source_sha256": "2ca780ce3e77229a45496b878d0889235388283e7c6933f33ab53d0a9924d89a",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_sbsat.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..2057246f7f25fc8c8ddeac7683d097af7368d5c2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c6e3d21701f1f2e0bb79c3bd21dfd319dc27d206a74b2eb7ffa36852ac8880d8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:55:51+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_sbsat.pkl",
+    "source_sha256": "6895597249b02a6136e26709ec8ff665901a0b942f01d53dfc8ca3824a1b065a",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_story_sbsat.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae450e78e840d7c54d5d9ad5b8f393a9dd483e13
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_story_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3493f723f7b17ffef0df50046f43fb9add628af328fe25c49bc46ce7928f3b87",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:24:06+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_responsivity_N512_s0_tokens_story_sbsat.pkl",
+    "source_sha256": "16707f0ce2aa22f79ebd515535e01fd63d4ba07921594dde61367060fd81ea2c",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_celer.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec965efdb2c27c79a99acd965bf87cf62a1f39b6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_celer.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c26e750a44fd28563bce0f935f1c799a7dc42b1268ae2f34192f8b6de682cb0e",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:33:21+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_copco.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..ceda10c4585d9061632983da0a4be8b76d24121b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_copco.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0c6484856ef76ab6624f983e86644729636352f51bc896b4a902838182c8154c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:28:34+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_dundee.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..6921671a9442e12b543a2f8f7389be8ddf35a438
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_dundee.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "832ab7a7047c77ce5193b676c2d235c27a8a5c4695499223facd337cb1ba9ef9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:35:31+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_geco.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..51727f5c5bca47df5480e596269b762dfd15d761
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_geco.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "22b24b3f93dea7eea54420c6d6e85ebaeeeb247153e87f28b40f744cd77287b6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:46:43+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_de.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd351718974a21b009539833b009bef5bf2617ba
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_de.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fb3781b289731e0ed43642be133f15727fe10379220c756b461bfdb4f4b3119a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:57:11+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_ee.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..caa8b4948c913b1eae5b8ab5e242b8ea6944d69c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_ee.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d18d53fa2cf026f9dc7b0827c590e2e8f362cfc29b7b1d44ab8601b25b7d8678",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:59:21+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_fi.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..5997245031eb40925003c5486fa253bc05b61e9d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_fi.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "841784869073c6bc28eb919f46206d3f4f63daa5618363fe96eeb489f70dd4be",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:01:34+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_gr.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf6d513a2cfad23072100fda272eec8575c5e8b2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_gr.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2ac8d55d4c889442acee449931c3aa7ec39b068d37dfe2b618a8f2ac25c74bac",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:03:31+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_it.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..6da75b1ae04f1ee57cd6d042bcdd0fa2b0a3d014
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_it.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "08c2d937b40281818d6cba824edfb2b0439a1613fba4080cdae64a508722f4ea",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:05:20+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_ko.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..591677e7d1445c0b2a69ad5d9064b46fbc67d06d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_ko.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "16091d1a0e32013a90470a7e17acb7f7925ea42d78ee3a591ad878a7b401e02b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:07:14+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_tr.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..d4e04f54280f189c6fc32bfde32f89d3f7b5013d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_meco_tr.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c875178a11dc511eace562c6ddbdcf183b778ee6ff0f056ed4c10e6d50675918",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:10:54+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_natural_stories.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c607eefa30a95c2fc01041398e150cf157e8c5a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_natural_stories.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "143f17ac8700c71e8793dbdeb2079053db0b3b3334b3566f0215b730c3f223a9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:13:15+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_onestop.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1754f6f0363ba260e1b3210f2cb7c9a6365b367
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_onestop.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "041fd7220e5750a3da00564de0895e017e2bd1803169bf15dee33fe683bbfc4c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:19:26+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_potec.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..80a89d0a2533bb5eeb7fd30153082e5eefaef9a0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_potec.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "69015cc48e26d81676a78b631f1a7d8f4b933de537583077c97fd485c8fd3bf5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:20:50+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_provo.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..00ad12120e315853f501a2997c796a4113190b44
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_provo.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2ca780ce3e77229a45496b878d0889235388283e7c6933f33ab53d0a9924d89a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:21:40+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_sbsat.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..9acd37ceca04ed9ce9b7a96aba5b21449b6ce408
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_sbsat.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6895597249b02a6136e26709ec8ff665901a0b942f01d53dfc8ca3824a1b065a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:55:50+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_story_sbsat.json b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..8888301ebd457beb3ade77323296703fde49314f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_responsivity_N512_s0_tokens_story_sbsat.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "16707f0ce2aa22f79ebd515535e01fd63d4ba07921594dde61367060fd81ea2c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:23:58+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_celer.json b/manifests/ai-forever_mGPT_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e510f2ad007f8b4879a8abc6b497fb59dfa39c4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_celer.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "37d9ac58db7650ed499d458314b12c6e1eae2762ec4aa4aab784de8d144d3d9b",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:21:15+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_celer.pkl",
+    "source_sha256": "aa413119429f55400b7a5e9f7477b8bc7c46f40ce3fd32b147a7245ab7cba162",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_copco.json b/manifests/ai-forever_mGPT_surprisal_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..c07c9b143d1001382447654ca9464441610f5eaa
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_copco.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "77bd80b0e1c2d3b1a010f27080853aa78c70ac8763f443729f320b84f3897dcf",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:06+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_copco.pkl",
+    "source_sha256": "865a847219d9237b1b361cbf6b7cbd4782a8a13bc505216b1efd3736c3d2fa19",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_dundee.json b/manifests/ai-forever_mGPT_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..5c1383a7822bcf389676689b00495320ad6549e7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_dundee.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6e0ec583cb5bcd357995ce23bd3367044f4b4d688ce8ed1fbcfdce46c727543e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:32+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_dundee.pkl",
+    "source_sha256": "aba84c23503db1ecbe1a7a2a5ef1a095f101382149446a47d4cc3d753d05552a",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_federmeier2007_n400.json b/manifests/ai-forever_mGPT_surprisal_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..696714bb6e4844b406e28fb6d2eee2d2fbff7a53
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_federmeier2007_n400.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f7cec2f85c59c7d2eebb6c3be3b186174d4b68caad993298856082f9e5297156",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:25+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_federmeier2007_n400.pkl",
+    "source_sha256": "88dc774d02e12646dea1581903eacc9a8d52a1e0f4c3230c67bfb6a141fad8e6",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_geco.json b/manifests/ai-forever_mGPT_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..8073b0d51a4892eb5f3a9811083dfaede2aa3580
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_geco.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1265ed24e3f96dab7d404b41933040dac0a16deecf40c05cd0b84ca71d76a79e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:12+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_geco.pkl",
+    "source_sha256": "a2785503ae754e34ab34cd106b5b88248da0088d12b69ac4243471f63a48ead2",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_de.json b/manifests/ai-forever_mGPT_surprisal_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..a159467b067ca65e64b745b6a0fc4c522a74df29
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_de.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ed6823eb8c09c833edafcb690382a4437706383f54c708f98b22b1daf0d3c3cd",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:25+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_de.pkl",
+    "source_sha256": "8876ddc1e8f4589bd69311162e5ad9eb71acbd9b6067bddcf6f66c72f40a238b",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_ee.json b/manifests/ai-forever_mGPT_surprisal_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..fa75afe40af7c3651cc82cac7789b232cbea2466
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_ee.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d74387e848136ce9b91b087c59226e325ff5984d2e9c9e22974b8813d4e56f00",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:29+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_ee.pkl",
+    "source_sha256": "567a0e82e6016bf5e05248d0debb561e8f07f786c0c184b9e5b700dac4043172",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_fi.json b/manifests/ai-forever_mGPT_surprisal_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..d099bc3b0a9c12a76b9e3de51a8750e702622a36
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_fi.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f4b4b941fe2a527f419fef33d91f5345a51a910c4a709a70460e781a7be17727",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:42+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_fi.pkl",
+    "source_sha256": "c104ba795a12cd2fe8c678fd47c88810debad8abdd69d4ba27ecc1c2f78f6128",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_gr.json b/manifests/ai-forever_mGPT_surprisal_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..b324a2266d74a4d285a49d6dd6b360aacdd796b5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_gr.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4b1d8e085a6602959ac99711558537897061f52b60873fa1fbb297129c090aa1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:50+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_gr.pkl",
+    "source_sha256": "fe0f1f4e585a24fee349790d21cf03904f6432f357f09759e41b1d5983e81569",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_it.json b/manifests/ai-forever_mGPT_surprisal_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..26d3bc96f9ae35c4bb7a4bb9e56d956c4985ed83
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_it.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "325434e76593d5ecf5f32655ee49d0625e8191f6b9bbba080692cc84fcfde11d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:02+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_it.pkl",
+    "source_sha256": "2d3396acba52f18c3753e7ff838ce6b1fb440ec825aeafb3e5f7631c43a55ee2",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_ko.json b/manifests/ai-forever_mGPT_surprisal_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..912d14bb9c917a9085d365ec9fcf793546c51c4d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_ko.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "eab4309fd6935abb9e4d69b628d19a24133d8e1ec1462599ca5179209dee30b6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:04+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_ko.pkl",
+    "source_sha256": "d71431a9593fbf7bf9be33d86d99132a7564d59e97749dd336fd2db015c3d207",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1863aec2df70fc8a2c26e9c33144a93c2e0884a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_ba.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "41f74da23adb23aae29db1dc5ed453a508bb21ab72ae600d26d0e23a49bf4a54",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:08+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ba.pkl",
+    "source_sha256": "dadf7d24d08274f876417fe25f86b27ac275116a6c25cc2f5a91aac5a47c9825",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..d778f0b438617f40a746253c6b66141e5c572312
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_bp.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0257ad37e944a487ab5272a6f4701bf74b86237537653e30ddf73fc294449004",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:15+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_l1_w2_bp.pkl",
+    "source_sha256": "7511b14a5707b544a4c9cadf8484679bb35845219cce7939c087583a844b196d",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ed29eaa2f80f4bbcfe32a3110c5a9a759d034b3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_ge_po.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ee7731c132c265ff92652018cf644fc6e42150105cb80a42ec32486bd8eb7ecb",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:22+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ge_po.pkl",
+    "source_sha256": "76d067038d7e65d7c16fd064ff350b626dc895e8f0afbe107f9d70f1225a1066",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..d1b3f689e9d9440aba57dd7bdd8a13a30ead2407
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_ge_zu.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cbda22bdff9a586bdef8894612752ba7a68b4ed884ccc5459c10e0b3916397c1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:24+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ge_zu.pkl",
+    "source_sha256": "510e0d0d9971819c5966472a5866b87175d708e0943b4924c86385e9a4eb3704",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..5df8b9d1bd00773de56b6d06b46679b3ac3b6421
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7020f011aa88472d389b2a2a19348b899286761bba0837f376c07a94457af516",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:30+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_l1_w2_hi_iiith.pkl",
+    "source_sha256": "1b0fab651f74f0f571227bc768e4e6cd49dff42a5f7e8530cfef1caca927c633",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2b9a55d58a37dcf2e2909ce7a79ef9b0f39b45d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0447f2bbcccadae6784f26dc1973b29f382a270299ed0f7b2eeb16604e1985e1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:34+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_l1_w2_hi_iitk.pkl",
+    "source_sha256": "7375325d04955fc7ea18509aef5a96ccc9ffbf363556d64ab2f95070695699d7",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..c5c0effbc12e7aeed06e50dc9b09c3bb59cd556a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_l1_w2_tr.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "be7b7d4acf53cf6bc90aad577b58c9e935a7b5f9863fd7180c16e4c0563a74c9",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:26+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_l1_w2_tr.pkl",
+    "source_sha256": "a99d1a7774a68b6c32defd3a0762983b64bad30392f1cb2517813023d1e7111d",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_l2_w2.json b/manifests/ai-forever_mGPT_surprisal_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..08bd6963f4963709c8b0d0ade3fb53bd34c33667
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_l2_w2.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7ce9221310fc6daa35bb7640b0582f3eb83faa62f41739b4c212bd3b222b0fd5",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:27+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_l2_w2.pkl",
+    "source_sha256": "bffb661d5a409cd658434e6841be942c096ab679b0301cff3a6f75bc8c74e5d8",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_meco_tr.json b/manifests/ai-forever_mGPT_surprisal_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..72bd31458ba957dfafc64134c1861d6ac186cd87
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_meco_tr.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "721193fa609b54e5914ef25b65ddae8d6658669a12b595b52ef0afff7722381b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:33+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_meco_tr.pkl",
+    "source_sha256": "6c1c7a3c7af3feebecf3d17170094e299e53396d00d0e3f721ddc725248d0f03",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_natural_stories.json b/manifests/ai-forever_mGPT_surprisal_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc7a6cec6dca62f17bbfedd260b47b1a85c298c0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_natural_stories.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bc4ed0e095dc30b6fbafe9a931ccb04e2a20b01ef03dc3e944b29b9ca1d48062",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:41+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_natural_stories.pkl",
+    "source_sha256": "08e10211d1522e958154ececafa04c5250f9a4664593db65561cc45adda567b8",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_natural_stories_maze.json b/manifests/ai-forever_mGPT_surprisal_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a3e4924fbd1ada8d454f1143df922d84b412d64
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_natural_stories_maze.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7f1ec292d51bccc1f3230f4bf983c37414be681f0aa186738e834d6c16a93e12",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:49+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_natural_stories_maze.pkl",
+    "source_sha256": "df09065238341db7c9829a968ba0a56c3e80e60efc9c050043a27b19c4f4dc81",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_meco_de.json b/manifests/ai-forever_mGPT_surprisal_nobos_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..f26fa108bfc17a83671bd4b276e81e7033610bdc
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_meco_de.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ac084566ede722fac6cd18cb26a8234d1f326df22062902403dd5276eb172080",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:26+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_meco_de.pkl",
+    "source_sha256": "d16d5b3c0acf8185698416d10f5544208c875a17bc4bdf5a6420c384ba04706b",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_meco_ee.json b/manifests/ai-forever_mGPT_surprisal_nobos_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..34e75f402495710394d0b1de0277102fdfd820c5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_meco_ee.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "122155362c0bd669616478e0484c9fb08a33edb4615f99cd518126bc4284f13c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:31+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_meco_ee.pkl",
+    "source_sha256": "b8488b58e735403eb8e62a6e71bbc3ea152fb9085abe62d75864411df30d15e1",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_meco_fi.json b/manifests/ai-forever_mGPT_surprisal_nobos_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..e1e83854da59407c309d2c30b601c6f586d75047
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_meco_fi.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "db396f5a40ae0893ca5e1c9ae4c8611a0466e29c523d36db337a69058856cbc2",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:43+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_meco_fi.pkl",
+    "source_sha256": "16937b6ff2eea0bc3f7ff74f37f4727c52eddac6583d94a4b7121d022722f5d7",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_meco_gr.json b/manifests/ai-forever_mGPT_surprisal_nobos_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee93f71400984e26cdde85ac07c495865613dd1a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_meco_gr.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "700d3b817c8867332da059ed2854883d7c7b84eac1ab7f4241950dc8dac91cac",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:49+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_meco_gr.pkl",
+    "source_sha256": "159f5e1c1c528a97ed3e28a34b1e46c180b8afcfa3184b4c695d3ecab176216a",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_meco_it.json b/manifests/ai-forever_mGPT_surprisal_nobos_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..23484a738860033fd3ee05da34bf847adf9ccd9c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_meco_it.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "140b5e97c8acc4f68fec9d125bdee6a6b3c1c5ac3ded051c2d9fdd91c5147efd",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:07+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_meco_it.pkl",
+    "source_sha256": "118097e7a112f9c6097a4817452e6ddf7047be11ffc73dddf1e8373c4948094e",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_meco_ko.json b/manifests/ai-forever_mGPT_surprisal_nobos_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..c81e742d9c05ba5a881ad55e559eab700b0a702d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_meco_ko.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7a2ba21cf8b26ccf01a9029fe63546ae89807168e2fcd853b2455aade3b51e9d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:05+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_meco_ko.pkl",
+    "source_sha256": "447d9c94ecfa8395c3a5ec4a9f5fe8756872a24c62de073b48b8906fb1a67835",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_meco_tr.json b/manifests/ai-forever_mGPT_surprisal_nobos_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..f973bcc04b097141255c78931699d22b1ac85110
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_meco_tr.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "18296d4db09d4a41f28f926a1c0062a42b21e1ae881222a00b96c6751e8a0bf9",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:39+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_meco_tr.pkl",
+    "source_sha256": "163809f718d8ff25e7112fd74be52c42b320aa0038dfa9448c6e8e7a1cd1b5bb",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_de.json b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba1de5bb20f8270e3fb348c43b7a58e98fd59edf
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_de.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "8a91f15c9c999ca3d2b81d5f0f9033275054cffbf4dac78e99339230ee70057d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7077f7e622ed47bdcde4cb35f198f0752b5d5c4908f6ca6955be3966735ad41c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:34+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_de.pkl",
+    "source_sha256": "beac50fa0c7f7bababea37555e8be3d90171e85c0df18a278785e4f900da0f0b",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_ee.json b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..b13b9417ae65ff0615df345ae81ec50fa471d0d8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_ee.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a3bd7f10daded68cea3728334385d0638ded96ba61df00cc85eb4c98848aaaf5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3cbb135cacb0d5dcc6591e28bd9ec0b1f993d7b5377a2031e0546b338d9b159c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:47+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_ee.pkl",
+    "source_sha256": "974ddf651d97eadc7753877219960f64fd02ad8f14ad252261d6d27abf2924e1",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_fi.json b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..30081ffffb69a3b013189764e89190c7a94b9c53
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_fi.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4ea11b1a9e4318635b18087b3844bd3cfa3891f932742f2a3e93c0887c79b57a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "15382bb4cd7a2318926fc37e047cc15efee51610bf08fcc88e55425a4747e68b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:55+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_fi.pkl",
+    "source_sha256": "793bdf6ff79d8905f1198003ee2004f5c2e23da8fb4ea197c7629c7d497140c8",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_gr.json b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c9d3a0040119ce68fd7e06e2ff0ab3c155ed510
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_gr.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3fea08cc811f5cd24cbe536ebc4360c1a6dabad4819c7515a4cd6b68ca8155d2",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "111ddde1234650d05ca74028cf7e7203021a4f11cddea963b315f55460cb6189",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:01+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_gr.pkl",
+    "source_sha256": "272d8413a6ca076e9373f60426249d9f97b7dcee8e684aa6aab3948e33086f8c",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_it.json b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..c19122a6caaeb383b939e75e1e70991162981956
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_it.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "6da2884eee23d85230abb29366689af0289e5e99fc41fecba935ccd3ed55b148",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7e87769108fae059f3d2083b16dd0e1d65abdb123db157669c595f195f8a6725",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:15+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_it.pkl",
+    "source_sha256": "02a6bed9cc1fc92de0529d128cf895be9288bf14848fb7d378a461dc7217b654",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_ko.json b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..27521e8e65da393f52d219df4235ccdf7cb20da5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_ko.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "8a309737fd1a038f8efc74d3e3f6aa24d3415f1e8e7aca2ee3d39feccef6b720",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3933666787644d55a18d0f5aae34bef875de9c70efe5106f9fc7b01d7e199b9a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:13+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_ko.pkl",
+    "source_sha256": "f2667933e5b17acbd17ac0cae7916531cd2e4278a1f2fb6ad7573f3da0526ef5",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_tr.json b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc123759610408c0f69a1eacc06828910ffb414d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_nobos_sent_meco_tr.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bc412857c22dd42bc7497421e05d56dccb15e61a21e1eb9925554ec022807feb",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "814a34634e655873bfb254424172c94fc73ad27dbecc27cfdef5f8d9df47ca48",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:51+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_tr.pkl",
+    "source_sha256": "b688cc31fd5d18c56759bcd86fd724d604be5859bc94940306bd34f6e4ddb54c",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_onestop.json b/manifests/ai-forever_mGPT_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d196106bdcbf3469a54b14e4d2c1bd4e906b427
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_onestop.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7c5a84a750fa123613b6dba195d082d2e886217f888338e6d11a1aa68e22439f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:30+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_onestop.pkl",
+    "source_sha256": "3f413944ce4ae3c19a2db660b6bbe441d219498387f4218dfd015b53b496b5a3",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_potec.json b/manifests/ai-forever_mGPT_surprisal_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..e321043e90e6a30d0c24a727e73b2d2c0b374e0c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_potec.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4c8947247e79fa9ecea43d5eca13d3aa38d12c87d410d3d95747facdfc790777",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:00+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_potec.pkl",
+    "source_sha256": "9817f7a31443f98f942fb3bb5a1a489166767301ff115e61bfdc083495555fa8",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_provo.json b/manifests/ai-forever_mGPT_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8ee283d1a28e5339ce9ac1aab8a354a0a13263a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_provo.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "10dd60dc9136ffa18a7cc99ae0f8edebb5e585e25fc758d2f1c7c63d97ccc390",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:57+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_provo.pkl",
+    "source_sha256": "04092435e538425c8c8fc22a4c481b94888bbf2b5849211f93091a63a6a9cae6",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_sbsat.json b/manifests/ai-forever_mGPT_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..8ede8deb3d16d67aa9152cec28e8b2ff09b49d67
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_sbsat.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8fdb51d3b2320c239918c34ea2602bb911be044016f548471b9d44427c436175",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:28:01+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "5d238dea31b931455d1952e007b44e86ea128a995419f55b3a6e7759e59bc196",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_story_sbsat.json b/manifests/ai-forever_mGPT_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..02369c59abd2f7ed983f13ad3375823a0594d7e9
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_story_sbsat.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "16eee37488a93d2486ce86701e5b9a5febc527bbc6007c04ed45073f72843300",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:46+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "a9107effb9fb0026ffb2765c172997e858910a69ca25d41e9907c0624c6891f2",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_surprisal_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..51e10a55d50a6cfdd2374f295b19998b987e7c29
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "84652a58692d158a1256a747eb7a1b573966e4ad0d988cc8c8544e77366d30bb",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:50+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_szewczyk_federmeier2022_n400.pkl",
+    "source_sha256": "613f6154d1885d1b570ce950d7a53560d3153c04edc0a3a1643ad8de62471006",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_szewczyk_n400.json b/manifests/ai-forever_mGPT_surprisal_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..58561df698064f05d8c08c593ea672df507b895d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_szewczyk_n400.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3293c6727290c0ce18a5db2b737b92f46534562182f52e691b84460f66fe1af4",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:13+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_szewczyk_n400.pkl",
+    "source_sha256": "2f80bc146314863925b9863fa3c9e44f8e8ad00504f040bddd41725350a37086",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_celer.json b/manifests/ai-forever_mGPT_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e4fd2ee2d27ae15a0adb75d1010bead6ff9fb86
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_celer.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "aa413119429f55400b7a5e9f7477b8bc7c46f40ce3fd32b147a7245ab7cba162",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:20:54+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_copco.json b/manifests/ai-forever_mGPT_surprisal_tokens_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f77a446d9646a2b57a46ac5fe16c60cf1e0afc7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_copco.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "865a847219d9237b1b361cbf6b7cbd4782a8a13bc505216b1efd3736c3d2fa19",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:04+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_dundee.json b/manifests/ai-forever_mGPT_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..a110e76ada63fb096a88570a68c9afacfb3648b0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_dundee.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "aba84c23503db1ecbe1a7a2a5ef1a095f101382149446a47d4cc3d753d05552a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:30+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_federmeier2007_n400.json b/manifests/ai-forever_mGPT_surprisal_tokens_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f1bb193bc78a8098eabef76972771ce6f7824a5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_federmeier2007_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "88dc774d02e12646dea1581903eacc9a8d52a1e0f4c3230c67bfb6a141fad8e6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:25+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_geco.json b/manifests/ai-forever_mGPT_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..11f5cab9a63b77eb05e72de562da420d48d8ba35
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_geco.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a2785503ae754e34ab34cd106b5b88248da0088d12b69ac4243471f63a48ead2",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:10+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_de.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..58d9c4a22a6643dea87089bbc2e9ec18c70c7eac
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_de.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8876ddc1e8f4589bd69311162e5ad9eb71acbd9b6067bddcf6f66c72f40a238b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:25+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_ee.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..8160dbdd4563d809feda68c097da4d11db13c2b6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_ee.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "567a0e82e6016bf5e05248d0debb561e8f07f786c0c184b9e5b700dac4043172",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_fi.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..91cc2d62d7f759c45474c48e1f2b5367de3fd955
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_fi.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c104ba795a12cd2fe8c678fd47c88810debad8abdd69d4ba27ecc1c2f78f6128",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:42+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_gr.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a8f22ccce302c1997f4253726c7f7d76c0571b6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_gr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fe0f1f4e585a24fee349790d21cf03904f6432f357f09759e41b1d5983e81569",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:50+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_it.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..4de67c9c458243765eebef52754ba7c58da64ce2
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_it.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2d3396acba52f18c3753e7ff838ce6b1fb440ec825aeafb3e5f7631c43a55ee2",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:01+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_ko.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..24331072c584a4eea15ec2293a169f8751edb639
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_ko.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d71431a9593fbf7bf9be33d86d99132a7564d59e97749dd336fd2db015c3d207",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:03+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..cbf2977c47bd25dfac6fb7fbc7c42ee3c0afb617
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ba.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "dadf7d24d08274f876417fe25f86b27ac275116a6c25cc2f5a91aac5a47c9825",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:07+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..46867b2ac0ea4ef60776b38ef7c88b77b28ce462
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_bp.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7511b14a5707b544a4c9cadf8484679bb35845219cce7939c087583a844b196d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:15+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5bd33cd469f96694a27b35cd7b4146259a8982a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ge_po.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "76d067038d7e65d7c16fd064ff350b626dc895e8f0afbe107f9d70f1225a1066",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:21+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..04364f2af8bdd85f74a3de01a9b09c7f2a979b59
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_ge_zu.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "510e0d0d9971819c5966472a5866b87175d708e0943b4924c86385e9a4eb3704",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:23+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..90a1bd08c12048c1dca4a4e2a76e6f877a8e5d3d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1b0fab651f74f0f571227bc768e4e6cd49dff42a5f7e8530cfef1caca927c633",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:30+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..2229717a236aa6f88b59c4fa3e05f49231dbe703
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7375325d04955fc7ea18509aef5a96ccc9ffbf363556d64ab2f95070695699d7",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:34+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..52167a86966cc3d5c129d3313617a035209182ca
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l1_w2_tr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a99d1a7774a68b6c32defd3a0762983b64bad30392f1cb2517813023d1e7111d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:26+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_l2_w2.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..169e466ca2160bf9dbcf21053529e0083d71027f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_l2_w2.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bffb661d5a409cd658434e6841be942c096ab679b0301cff3a6f75bc8c74e5d8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:26+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_meco_tr.json b/manifests/ai-forever_mGPT_surprisal_tokens_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..96268865deb07102478e33d2c7fb02ac4b6c711e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_meco_tr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6c1c7a3c7af3feebecf3d17170094e299e53396d00d0e3f721ddc725248d0f03",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:32+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_natural_stories.json b/manifests/ai-forever_mGPT_surprisal_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf574c0cdb22ca2b1cec4231310429698b48a32d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_natural_stories.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "08e10211d1522e958154ececafa04c5250f9a4664593db65561cc45adda567b8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:40+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_natural_stories_maze.json b/manifests/ai-forever_mGPT_surprisal_tokens_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..39dff1c8d18816bb557016f291a3133846d12acc
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_natural_stories_maze.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "df09065238341db7c9829a968ba0a56c3e80e60efc9c050043a27b19c4f4dc81",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:48+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_de.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..4dbd1f2b5a8687a76718677ae0a2568664473d89
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_de.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d16d5b3c0acf8185698416d10f5544208c875a17bc4bdf5a6420c384ba04706b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:26+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_ee.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc4f3080f6c6729bcede6d84dda99e790d50d299
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_ee.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b8488b58e735403eb8e62a6e71bbc3ea152fb9085abe62d75864411df30d15e1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:30+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_fi.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..1860fc3dd8c815a74035e7d7b9736d0071b3c682
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_fi.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "16937b6ff2eea0bc3f7ff74f37f4727c52eddac6583d94a4b7121d022722f5d7",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:43+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_gr.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..bce5dc090995518b1f2bf44de45fe6da5c1a24a4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_gr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "159f5e1c1c528a97ed3e28a34b1e46c180b8afcfa3184b4c695d3ecab176216a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:48+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_it.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..14ad5b1ff5168fd42a219d1166b87be55362eb4b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_it.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "118097e7a112f9c6097a4817452e6ddf7047be11ffc73dddf1e8373c4948094e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:06+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_ko.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..b1fd089096e158a96e4d9b218fbb3e682157f12d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_ko.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "447d9c94ecfa8395c3a5ec4a9f5fe8756872a24c62de073b48b8906fb1a67835",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:04+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_tr.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..8fc5188438ce27974f2e309075d1a821bef065c5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_tr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "163809f718d8ff25e7112fd74be52c42b320aa0038dfa9448c6e8e7a1cd1b5bb",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:38+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_de.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..9fc1836ebc15f1f0703e7bf21c94db20ca866a2e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_de.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "8a91f15c9c999ca3d2b81d5f0f9033275054cffbf4dac78e99339230ee70057d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "beac50fa0c7f7bababea37555e8be3d90171e85c0df18a278785e4f900da0f0b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:34+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_ee.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..7593d19d2e54ad1e4f6d09e4cb4eabb035e34d2f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_ee.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a3bd7f10daded68cea3728334385d0638ded96ba61df00cc85eb4c98848aaaf5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "974ddf651d97eadc7753877219960f64fd02ad8f14ad252261d6d27abf2924e1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:46+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_fi.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..46b6e93ccd9fced07de4f1a930d2e68c31a52eba
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_fi.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4ea11b1a9e4318635b18087b3844bd3cfa3891f932742f2a3e93c0887c79b57a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "793bdf6ff79d8905f1198003ee2004f5c2e23da8fb4ea197c7629c7d497140c8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:37:55+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_gr.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..45a6c62b8b645eb413fbc29b3c9bfb1aef30295c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_gr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3fea08cc811f5cd24cbe536ebc4360c1a6dabad4819c7515a4cd6b68ca8155d2",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "272d8413a6ca076e9373f60426249d9f97b7dcee8e684aa6aab3948e33086f8c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:00+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_it.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d6541b9b5a3693dd2f71e648994bc989db0a9ae
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_it.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "6da2884eee23d85230abb29366689af0289e5e99fc41fecba935ccd3ed55b148",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "02a6bed9cc1fc92de0529d128cf895be9288bf14848fb7d378a461dc7217b654",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:14+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_ko.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..b1fa3da7bfb30fe9833a2ed772c2ddea220507d0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_ko.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "8a309737fd1a038f8efc74d3e3f6aa24d3415f1e8e7aca2ee3d39feccef6b720",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f2667933e5b17acbd17ac0cae7916531cd2e4278a1f2fb6ad7573f3da0526ef5",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_tr.json b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..868ba9cc4254e5d60e64d8560075ac55daeb0539
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_tr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": false,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "sentence",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bc412857c22dd42bc7497421e05d56dccb15e61a21e1eb9925554ec022807feb",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b688cc31fd5d18c56759bcd86fd724d604be5859bc94940306bd34f6e4ddb54c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:50+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_onestop.json b/manifests/ai-forever_mGPT_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..519ea25a67e9c00031afb517cf0c0aca27d3d197
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_onestop.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3f413944ce4ae3c19a2db660b6bbe441d219498387f4218dfd015b53b496b5a3",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_potec.json b/manifests/ai-forever_mGPT_surprisal_tokens_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..a296f82e2a100b01719c826b7d8536907ceec39d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_potec.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9817f7a31443f98f942fb3bb5a1a489166767301ff115e61bfdc083495555fa8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:59+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_provo.json b/manifests/ai-forever_mGPT_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab8cba71721e9cb4ab02ca5bf05db2199495ad1a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_provo.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "04092435e538425c8c8fc22a4c481b94888bbf2b5849211f93091a63a6a9cae6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:57+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_sbsat.json b/manifests/ai-forever_mGPT_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..814b625afa006d284bc0ae4b389175d0b2442dad
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_sbsat.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5d238dea31b931455d1952e007b44e86ea128a995419f55b3a6e7759e59bc196",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:27:48+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_story_sbsat.json b/manifests/ai-forever_mGPT_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..186e66bf295c86b5f8459f68c7fee6eed9e39530
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a9107effb9fb0026ffb2765c172997e858910a69ca25d41e9907c0624c6891f2",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:45+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_surprisal_tokens_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..382e643dba3782be901cd2682b510fc99e45ac22
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "613f6154d1885d1b570ce950d7a53560d3153c04edc0a3a1643ad8de62471006",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:50+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_szewczyk_n400.json b/manifests/ai-forever_mGPT_surprisal_tokens_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0be8ac66dae3caa5a962191e0cac5031a31fb75
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_szewczyk_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2f80bc146314863925b9863fa3c9e44f8e8ad00504f040bddd41725350a37086",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:13+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_surprisal_tokens_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..a823394d55d192c96228f89c819c8caf5c2cf263
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_wlotko_federmeier2012_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6584d33049e809c34a1d70916e664ba6bfa5a471213416bc19da3690a8120c62",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:53+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_tokens_zuco1.json b/manifests/ai-forever_mGPT_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..63554be685b8fe037ec61c55997a91cc1cd4fc1b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_tokens_zuco1.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b79708447ef778b848c1d914c516535e650301bb408a06d1d242480806c69cf0",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:15+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_surprisal_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..7600b2dfcc09efb53f7141344bed6e743f9722e0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_wlotko_federmeier2012_n400.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9190416369e050f8d16dd22b99b4b97cf84c7debefd78f0afc4311a4920cfb1a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:38:53+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_wlotko_federmeier2012_n400.pkl",
+    "source_sha256": "6584d33049e809c34a1d70916e664ba6bfa5a471213416bc19da3690a8120c62",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_surprisal_zuco1.json b/manifests/ai-forever_mGPT_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ea93ae271b072213680e123f391b763602a9419
--- /dev/null
+++ b/manifests/ai-forever_mGPT_surprisal_zuco1.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4ba7c9609f93ec0721db933e07b48db505b393e60673870801a5a35ca6538e64",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:16+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "b79708447ef778b848c1d914c516535e650301bb408a06d1d242480806c69cf0",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_celer.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..3020aff8228596c5a9f8b7065bf019a8d01fc09e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_celer.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "12c733d543f27a9fb09b4f82bd0482e7cde63bf6b0e2c125cdf189fc882b38c8",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:28:21+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_celer.pkl",
+    "source_sha256": "5e00d4babd8e7c50a2f46b906a4e044f8f896fbf3ee69abeb580fc1bcc117881",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_copco.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..bcbb905524c15f794d88132d5d4d0521ac0a14fe
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_copco.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "83617a88d13c9f039e73f41f8b2c135e1a2e673b2b7698fa26762122aa9dcee6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:12+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_copco.pkl",
+    "source_sha256": "8d8c7e05bbda59ac34a21c25124a48d98ce0ef20d21f659bc7d021930c3fceca",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_dundee.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..12512082c4b13d740de69557adf1a05642c71c3a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_dundee.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6875fda90d5c83b1cddda08866de72d3bea43ef9e75c61acd87b71e3e0dbf366",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:40+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_dundee.pkl",
+    "source_sha256": "4404e56603be8dcd786f3194eebd55d432925c01440f226036f9b60c42c7e51f",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_federmeier2007_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b21806d90b42c9a9a240686ec1bcdd815d4247f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_federmeier2007_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0e2ecc3491e4458b74e045d034ac4bec08fc91d3bb00188ddcb10cb9f573f0c2",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:15+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_federmeier2007_n400.pkl",
+    "source_sha256": "1270eabb9e650120da0edf2ea88fc22b828481a21dc0ae18f72cf1f19633bb9b",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_geco.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..8acb744010124e96e43db702e252b675abde2a90
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_geco.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "db7ca77ec179ca73817e273c72e55e425daae1ce2243130a5f65c899f46f9319",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:44:36+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_geco.pkl",
+    "source_sha256": "6971af4f05f03d8c15e98d2da7d582cc6dc0d557edc33bce0a35d08c2d89a8eb",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_de.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..e539664b06dae5bdef68e183e8747460297ad7fc
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_de.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e88e7cbbdf3e15b80db8037151ed4c3170c923f436bf56de83ee548ad813b490",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:02+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_de.pkl",
+    "source_sha256": "72df1f058cbfe85f58eca0eb041588f1e01cdd4cee9bc0651548509cf91e0017",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_ee.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..8ced4e73ffa562375cc85069b17116b83301ece3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_ee.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "aee89b0b929dea70e428e9d616cdd70d0549f5a632439fbbae2c0d5a7c71fc4e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:28+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_ee.pkl",
+    "source_sha256": "52598fb2ae77774fe2407692033f2c9b14f62e002f081bfa04e40e020b75e1f8",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_fi.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..d16dbe4d92971f50ccddd09ef34fb4d653136bb6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_fi.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1eb78302cd258e4f66d5d13089eef2838ceeec892c1ac7949669757f4d58c8ac",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:28+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_fi.pkl",
+    "source_sha256": "b776c2989cc1e259237aae3ddaf526f427ea88665d990f67f62f72206289d1fd",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_gr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..893139ec5ccfce1986c6b4a87df6893477509164
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_gr.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6259d8f4679d78ea35c9b45b8bdc758d60d42410becb326cdfb59f81d705f61c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:27+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_gr.pkl",
+    "source_sha256": "73b6bb39e1781b21a9e46e6b2bd14864250e2fc7caf9c8929e42f387a37e485d",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_it.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..1045db3192fbe32b146d2709d3ce47d42b5c66fc
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_it.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b85106d46c00d4d4a75896579679f7f930b48fba63abe31b628f62037733bde8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:33+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_it.pkl",
+    "source_sha256": "53d231c491e086bb02af10e10b719b0c120e0e4f10e281bd063a6a0e4fdfe621",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_ko.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b33eb9f66c191b668737ed2eba0e759675d653e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_ko.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a8a72196ec8b538ef5b8be2facfa889abc1d74b34f46e799df33621021f1bebe",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:32+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_ko.pkl",
+    "source_sha256": "49dc4e53a14043d044d56815534973a3834e6f2bb05f3fd0f5acddb825b92819",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..71398b8aef3c7bdfb5b508ddf9e173e3e608538c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ba.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b904dc80a04b786da40ab30d9c6d5f53659b0639bceeaef31e0bbccdbaa12668",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:46+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ba.pkl",
+    "source_sha256": "f40ca983543316aa58b6cd4bb6bf778e017fe027f569127c668b218c5191c66a",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..e81c787537721df6d07c1d49c369c7adc364290b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_bp.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "120a83a4e9a5c9d7381ba76a776f33a857ef59dc3ee9cd1d4f8750e32839cea8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:47+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_bp.pkl",
+    "source_sha256": "529d159b56daca6c79c130e9684ce39ab26c746c11170ed78619260878d49a92",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..16b7450a72d50b9329057a7f7338e6977058f819
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ge_po.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "78871a41081f40fa30d2a6246060275654cd61b460b3a72269be3d05073206c9",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:58+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ge_po.pkl",
+    "source_sha256": "da9876a1cc641a35daeb7db75f17ca916843dbbdf121e12e114671c8adbd58d3",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..102a6dad4d7f35297ccce24abb8e4ae98ba802d3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ge_zu.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2b3a4da2957e588af0456d7c1e42f90741dcdc59c8766969ea14c91073b476df",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:05+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ge_zu.pkl",
+    "source_sha256": "04c8d507662990b87875379dda0c28d07d337f497185d5b34435d89173068a0a",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..dad346735be71bf7fb933bb9a8c506a418b2094e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5f88c3ed98967d1ab656cc7b342ed245a473ab68a21e49ea4a26b65f8b2c570b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:16+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_hi_iiith.pkl",
+    "source_sha256": "09581397f20da4b5ee61e48fcb2be009e3d1df4f641aee27bee5be9425d322f3",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d9076a7030f037878cb80a0f3b4243041b0ea7d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "85e63c5aa6c276acc2bd160b90330660bf822eb05e1f51b07274ce15e0aa9eed",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:17+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_hi_iitk.pkl",
+    "source_sha256": "1920123475244038af69127a2d538e70b5dfbbad0538ed7635fde95b37c16e05",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..1def0be045e2f87c36571a1d2564f670eae70bc5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_ic.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "207c5e1af176af98a709ebdc8fdbbe170f6cedd459e0b20c3a078be5ed1d884e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:12+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ic.pkl",
+    "source_sha256": "06f77106c2580863b9ea4501de5cb10c16fecba59a3105eeaa87ded81cd2d91c",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..535af54dd22552794e1bfec7b790cbe85f27153d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l1_w2_tr.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ae1c4bb608786577786ba6a1737fc038b59647780d77a5af53d3d56aee75b0df",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:06+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_tr.pkl",
+    "source_sha256": "233823d9d84305cb5be3457bc7b11f0186acb881b208d3863a513e2a05c86a55",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l2_w2.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..594ead6da99cdbca40ac3e87f3b323936502ad53
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_l2_w2.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "82d2d0682d04a7038f7e3f5bb95ece1b45f90b0ad9ab3a80a574222bd63e35b2",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:04+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l2_w2.pkl",
+    "source_sha256": "8f2ffce397ba41b4bd4e233ddbe208a4b3411a1c0ca3483e7b3ca74b7f257c16",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_tr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..c22dd88e24de86f9e678aa065afcc6bbf4c83be4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_meco_tr.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7f13c11e9be851bef092cfbf46b4ab914d858305e1d0d4dfb9916d2a2d5c5ea1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:19+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_tr.pkl",
+    "source_sha256": "47d444fcd0b45e0038973d027d674d349f51e9ac45c66491de951be0921c3ee6",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_natural_stories.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..0902259cb0253ba29208f95f84ce4b665f5430be
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_natural_stories.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "46c8a91519763b8afc429945ff0dc944f4d8bbd4ba998ee8e8df989a0b90e495",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:22+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_natural_stories.pkl",
+    "source_sha256": "4a0f7a0132be57361d2f178db57fe9eb9aa2c81be274fd27a51a0910b2078ca0",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_natural_stories_maze.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..412073e231b14bde87dd2e09e6a158e452a45f63
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_natural_stories_maze.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "beafeaa76ef7877dac7badf624c81c309d13fc1aaba2b0c7e42f016c2e0c381f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:32+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_natural_stories_maze.pkl",
+    "source_sha256": "c0b209cc43dbe8d6514caf45de8939f60d5b7409c2e43ce3bc32d8bb672412a1",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_onestop.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..1de491b2b0b53d17b79f9bfb260aa0088aa1c57a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_onestop.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "645df0722be9dd31fa9483871021758022a848abf043845dc26d550d18577fd9",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:45+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_onestop.pkl",
+    "source_sha256": "703f81085fc905ce0aea9c15a7bb029525343b3b7115b037fea2bc6615030d54",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_potec.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..66375c64c812a73c930ac7b3073a30a29ff3704f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_potec.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "16d6233d18b499fcbc979a52a8142d8d2e5bfc50767757901f8a2c500d24ca34",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:01+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_potec.pkl",
+    "source_sha256": "6ac1bda7d82ce48a608d76f8b222edeeac3596d32186f2f5c3c63b000efb772b",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_provo.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..c35e72f4cd66d845374c57044b9caef6f35c2c12
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_provo.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "912a3ca4d58b1b0b98c4bb41cfdf612c72321869ea145f686f47df57d087caf7",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:56+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_provo.pkl",
+    "source_sha256": "43992a975537b4514eb0bf31932dfa6bd9bb8de71aecc108afacfdd4c05a6c31",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_sbsat.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..0d8ed7933c6df6429695e2044a59ec719594ef04
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_sbsat.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "992856803d825fe33774bed453e44f89e0e9af2441a70b9bb7ff6b94dfb9e88d",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:28:01+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_sbsat.pkl",
+    "source_sha256": "47f002df112e7700da9238188ead025beb9aa251856c8aebd1ef1f5734c3ae2e",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_story_sbsat.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f76cc977275b8aec168634c078374d14724374f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_story_sbsat.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6337b8c39e3be7d71daff95e0f8f47df96d1c2604ac971ee1ef8e0924411ad1c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:50+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_story_sbsat.pkl",
+    "source_sha256": "fb796f6e29557ef98bec4c510b6d1f73ec4da9f12b2192400e54c7e2bea0cb53",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..6276ec2b0ca2def6af8bbd00e60dae08245f0873
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "36ed9b2b0ac08669647ad4ef785a4906de3ea96ed28050e55242de2f15eb498f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:53+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_szewczyk_federmeier2022_n400.pkl",
+    "source_sha256": "aeb77b82f7d133f6df46a4be07759a786787bff0a08abd9d06713b9ec7658643",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_szewczyk_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a042d3f54c43aee8f083c117fd4628291991003
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_szewczyk_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b82a4cb0da4a0c0acf569bfe44b5e8cca25a9b576a1ce230602353b665f845cb",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:22+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_szewczyk_n400.pkl",
+    "source_sha256": "2494dd825c7e4f32df23b4fc8b777f573a09e269e24a7deb825200a9a6879f85",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_celer.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c93625010b910bfdfb9840315953b6a39011eaf
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_celer.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5e00d4babd8e7c50a2f46b906a4e044f8f896fbf3ee69abeb580fc1bcc117881",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:28:03+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_copco.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2c55dcb1a261d84ec9a41f3f0183f4b31503a93
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_copco.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8d8c7e05bbda59ac34a21c25124a48d98ce0ef20d21f659bc7d021930c3fceca",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:11+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_dundee.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..f20fdf6c7d8d4c78696211260ed94f8998cfb92b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_dundee.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4404e56603be8dcd786f3194eebd55d432925c01440f226036f9b60c42c7e51f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:38+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_federmeier2007_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c9af598489f506a0ef8a492d25e888f5c9ca0e3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_federmeier2007_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1270eabb9e650120da0edf2ea88fc22b828481a21dc0ae18f72cf1f19633bb9b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:15+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_geco.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..8727a39c1641a5e6426508f230b56fc9787de411
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_geco.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6971af4f05f03d8c15e98d2da7d582cc6dc0d557edc33bce0a35d08c2d89a8eb",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:44:34+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_de.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..342768e3f8bff2adde1f27c2ea5a252f65db82f4
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_de.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "72df1f058cbfe85f58eca0eb041588f1e01cdd4cee9bc0651548509cf91e0017",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:55+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_ee.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..f86f921b91f43137653ecf01bbceabf89c7410c1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_ee.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "52598fb2ae77774fe2407692033f2c9b14f62e002f081bfa04e40e020b75e1f8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:27+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_fi.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..c737a8449bce426f6756ed355d850cae32c37518
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_fi.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b776c2989cc1e259237aae3ddaf526f427ea88665d990f67f62f72206289d1fd",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:27+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_gr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..a04d68b3a22cde5a7898c363569c7eb99d8c1152
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_gr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "73b6bb39e1781b21a9e46e6b2bd14864250e2fc7caf9c8929e42f387a37e485d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:26+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_it.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..590085cc48c85f52a53314a61658427963fcd31e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_it.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "53d231c491e086bb02af10e10b719b0c120e0e4f10e281bd063a6a0e4fdfe621",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:32+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_ko.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..554f97a80dca72c59b24558517830479e76029db
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_ko.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "49dc4e53a14043d044d56815534973a3834e6f2bb05f3fd0f5acddb825b92819",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:32+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..97e427fdd3a3ff6088e4533ee04ad368c5bc8ec5
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ba.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f40ca983543316aa58b6cd4bb6bf778e017fe027f569127c668b218c5191c66a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:45+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..64319bc5087fefda147d62672b45af5be251e667
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_bp.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "529d159b56daca6c79c130e9684ce39ab26c746c11170ed78619260878d49a92",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:46+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ch_s.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ch_s.json
new file mode 100644
index 0000000000000000000000000000000000000000..b68b738611edf791fb754d548a4098a99a31b1f8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ch_s.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9f27685ea775f0843723c5aa8c5b226b7d3420ccb213afd1eb8198439dcdb6a3",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fd98c53b555d30219091f32da3d0c34e033d2a8e4cea32ee304a90a8476e19a2",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:19+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ch_t.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ch_t.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5876ac5d75efa6f474154ff9a2878d95691c369
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ch_t.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "866884887c127db9b93a7cf7e4deb154d4c2cba960aa58ede0d05493044e3877",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "299f442e22f9e23cec4bf7f8bc39a7e3f952779f7e1bd8f7f80fe9a4f6b04b35",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:20+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e36ae744396a2c38db173aaf2d0baa667256856
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ge_po.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "da9876a1cc641a35daeb7db75f17ca916843dbbdf121e12e114671c8adbd58d3",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:58+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..bacce117eea700e830d6fb92a43bb90c5f1e61d1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ge_zu.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "04c8d507662990b87875379dda0c28d07d337f497185d5b34435d89173068a0a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:04+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..96b8c95b2a8f19eba082225b643c0b94c721a622
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "09581397f20da4b5ee61e48fcb2be009e3d1df4f641aee27bee5be9425d322f3",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:15+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..ecd9cd98779cc2104e09b390565a65db8e601c94
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1920123475244038af69127a2d538e70b5dfbbad0538ed7635fde95b37c16e05",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:16+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..5639b7acfffbf2a43021cd8c784f6b21f6b76029
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_ic.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "06f77106c2580863b9ea4501de5cb10c16fecba59a3105eeaa87ded81cd2d91c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:11+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..b39b2563cdd5b0bf1c50b13b4f67433add36653f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l1_w2_tr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "233823d9d84305cb5be3457bc7b11f0186acb881b208d3863a513e2a05c86a55",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:06+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l2_w2.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dad9dcad14285d46430518feddcc96ce4689681
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_l2_w2.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8f2ffce397ba41b4bd4e233ddbe208a4b3411a1c0ca3483e7b3ca74b7f257c16",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:04+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_tr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..f77a7e83e377e26408bdac9521705bd16ecae963
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_meco_tr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "47d444fcd0b45e0038973d027d674d349f51e9ac45c66491de951be0921c3ee6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:19+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_natural_stories.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ed8ebeeaedb828108750a3e1a4b64956b3c1397
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_natural_stories.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4a0f7a0132be57361d2f178db57fe9eb9aa2c81be274fd27a51a0910b2078ca0",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:21+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_natural_stories_maze.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..935c2564ec877c7b6722f19e2cd38752aed844cd
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_natural_stories_maze.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c0b209cc43dbe8d6514caf45de8939f60d5b7409c2e43ce3bc32d8bb672412a1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:31+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_onestop.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd8ccf07f14c55b23beb157efd90fda20ebc09c1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_onestop.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "703f81085fc905ce0aea9c15a7bb029525343b3b7115b037fea2bc6615030d54",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:43+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_potec.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef9cc80176bd82acacb743603819a8e90160cb29
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_potec.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6ac1bda7d82ce48a608d76f8b222edeeac3596d32186f2f5c3c63b000efb772b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:00+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_provo.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e4283053bb311d04e15d25316507236aa617f32
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_provo.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "43992a975537b4514eb0bf31932dfa6bd9bb8de71aecc108afacfdd4c05a6c31",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:56+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_sbsat.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..c4dd406e7c68f4355cd6c18f88a8b4ef1642a59f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_sbsat.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "47f002df112e7700da9238188ead025beb9aa251856c8aebd1ef1f5734c3ae2e",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:27:48+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_story_sbsat.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..ed5c991aa9ddc4d0290071fa036f0044cbc25f42
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_story_sbsat.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fb796f6e29557ef98bec4c510b6d1f73ec4da9f12b2192400e54c7e2bea0cb53",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:50+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..13355141dfa016e44daa066f56e7313c5bf56cea
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "aeb77b82f7d133f6df46a4be07759a786787bff0a08abd9d06713b9ec7658643",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:53+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_szewczyk_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..890380da7a04921026da867a29e536534b156f96
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_szewczyk_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2494dd825c7e4f32df23b4fc8b777f573a09e269e24a7deb825200a9a6879f85",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:21+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..623dfc3414466ec25f2418985a6414be9b756d39
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_wlotko_federmeier2012_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e425e0e8214b60fdf5f2216762694cf521eb830f3f6e0ec8b526ced85bb10467",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:45+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_zuco1.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..801778cae80fb44d963467de371226be83da0f5d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_zuco1.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7a8a8b06c3f6600c8f9ece3b4f99edcee62ee0a564f95e8636b5932a157cef43",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:10+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee7d584ac9a4eac983ed0e9d688b598b15107556
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bd6ec47b94c2f574e3b3b1ad757acef55d2de684a7b8cc0f8042a84d0fb0cbd4",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:46+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_wlotko_federmeier2012_n400.pkl",
+    "source_sha256": "e425e0e8214b60fdf5f2216762694cf521eb830f3f6e0ec8b526ced85bb10467",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_zuco1.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0d3a44b3f65b03de2892a79f25edffba4502ea6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a0.5_zuco1.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ad5cc961ff6fbd470c59de977fd10b2a97373c3ae4df57cb90d6e535bc7be6ab",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:11+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a0.5_tokens_zuco1.pkl",
+    "source_sha256": "7a8a8b06c3f6600c8f9ece3b4f99edcee62ee0a564f95e8636b5932a157cef43",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_celer.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b811f3e3fae201123fd092752fa1c9365076cf3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_celer.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "08787bb4ae609d8a5c9d9a5b8e2ae1671ba6d10c21527b0f10a39b8c6851e925",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:29:39+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_celer.pkl",
+    "source_sha256": "837253e56e43f42826fc72235c50fd6824b45935ca39e4fec41f28cef5b6ecbb",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_copco.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..a81f3c39df6b2df2824b098a0f6f7b7a62c6b434
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_copco.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e38c4e588cd1099299e7680a50035824461a70bfd40169449bfd337bcaf66f0d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:14+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_copco.pkl",
+    "source_sha256": "8283a225a6d2e61acc23391739e5fb16421010cb30fcf256cb88f9612251b9f0",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_dundee.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..12635d86284e501b07f891d982b7657d51fa653c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_dundee.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "41657e9008ed78e271e2840ccf57330ce0e2b1dcc7f187cd6ac2e7d6ca240fcd",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:42+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_dundee.pkl",
+    "source_sha256": "43bc5d314aa12f860812618dd42f5cfb18e708340dadb197166568219f748af5",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_federmeier2007_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..27822883aad25a38ce4407239f39a15b61105bcf
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_federmeier2007_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ad7f85d19b6a9369a05e71e5bc5b46f034ae5b094c6702a6021140cf2eee4a29",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:26+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_federmeier2007_n400.pkl",
+    "source_sha256": "68155f1d7c674f2051b37c699fd25e94864ab1b817822be02577206a295a5af6",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_geco.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..75380abf6151ef7d67f390de3d3a5a63bd3fe16a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_geco.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2da21c8b9992cd50771061b61956048b26cea15a040e9a9ed7da481584ccf52f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:44:43+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_geco.pkl",
+    "source_sha256": "29d7db63d22b94de77859495cf2a12e609b86f056fe4f326e76ebc0933159ad8",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_de.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..59f749714445f983dbd40d912c8e84ee1e72de43
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_de.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "46e31b90b806a113390c651cf0ac68f9ab028e7572b2dcf04a1a48985ae274eb",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:27+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_de.pkl",
+    "source_sha256": "91502de01b59018c27660f925ce36596d7721e6e7eabf8b6802dcd271f6420b9",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_ee.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..53364f6d261b452a01b2f7f7beb598649a17f2ef
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_ee.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fd68c7d40addcd7212959c28afdb3bad53663b8cbb815962699f985f131bca9c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:27+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_ee.pkl",
+    "source_sha256": "937e42d90b1b395289109d00516700b41560042632e118d8f85d8edb1eae41fc",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_fi.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..3cfecf312385d743720cc5a3afeefa152266d716
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_fi.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "28bbe513ac925686aad88e073d606cff00098798afd302988a0c1d739b6e9ea7",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:28+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_fi.pkl",
+    "source_sha256": "f1a08b85575feb0d3790f1de25e6e2a61d05cad93969dc9bca3d3d3eac584145",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_gr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..772452df5beeb4c33618873a782bf71839f15469
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_gr.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "59e24048daad8796191333d763b52a4d9cd26b452861dac3b0f7392b5ee89c97",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:29+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_gr.pkl",
+    "source_sha256": "e2616e0270593c60866b80e53f827a72ee4e2b879119a69e48d745f6c018f365",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_it.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e1c900cb2ad10fd168c0b96db4c2a2ade0f3078
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_it.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "43f49d9658440aa39a0087d7ac7bce5aea267bd0e2715a1bedf1d3c1ac6ae3ec",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:37+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_it.pkl",
+    "source_sha256": "1041f9b8f8a4dab878158b7bd44340b4eda9353309ac4b0db5cfb2165296385c",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_ko.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..eea134a5d5e27a76989d4d734abd3900c30551b3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_ko.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3883b36c90d887aef64c3435bdc05c2e73bc84cedd16b6f579741fc6d851f35c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:34+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_ko.pkl",
+    "source_sha256": "e113dff38c1921b42126dd5e1d8a9d838b75570848a660bab6a9bf5b48637759",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..58045fce04815445a6ef5dae55dcba796439536d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ba.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fcfe84db6395bf4c39014055924b522cb1c3e1ca52c01a2f8a81c07041dfa972",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:29+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ba.pkl",
+    "source_sha256": "ff25dd55b9d50478ee2845ca3fbb583fca96e6b38b9d89badcf39e12eccb8773",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ad1859e435dfdb938636654b01a3bafab78b594
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_bp.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "33ecee10d62eb5e98138931f1ec2db313534c42fab53016c5df1540dd8f18087",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:52+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_bp.pkl",
+    "source_sha256": "3b05e04c33a619b72653f4d3f4d820ad26ae79b068e9b03d19d8f6d9d3542949",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0cab3c7d66f16678e8c01f7b7e0c5653f0e5260
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ge_po.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "235302e3c08238a69a1e0ac12d7817476c776d800de9c3cd58d01a37d4fb0746",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:04+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ge_po.pkl",
+    "source_sha256": "7cfa164b1ed7dd4879f092f386deea81f40d0ce3e83d14f5d9025816629b64a3",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e03086c2ea365a56b19b37926ccd1fead37bf8a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ge_zu.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ff1ef331ae6959b4db1099bec723d7f0d1284b447bac11e4b41b4487587e5981",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:07+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ge_zu.pkl",
+    "source_sha256": "93e3ee6dfc6619b1f49e538b7253307ad00be4eb71ca770b0301ea9368e09ec4",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..de5add22675f653ae78fa8b5311504126579c862
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "972b05aacbd0d522e789d3a880a242176589e1e43461a5d229844cba3e7581b8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:15+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_hi_iiith.pkl",
+    "source_sha256": "4b43959b5c1a5260a8213a3b7728e0a180178907106f48dd4e6d722e13871b89",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae2bd20514fbe81e5ce99f211501467da0d43ef1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c91d342b17cb06c9e34f45360e3ef26afb76d5768b89cd9b9ff07b8ff6b6c5e0",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:17+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_hi_iitk.pkl",
+    "source_sha256": "08a78df391c3ce52132337cf1dbb199a372ca591e16295ad02749477e58d55aa",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..b8c4a020a8d9786858e0322d9c7f33f31c5bae5d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_ic.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2f2f4a1622973d03e9ac9d9721240cf8e2435a3d498a2a973f904770fa1c1c73",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:13+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ic.pkl",
+    "source_sha256": "1403748a80372116299daa4de69ca92bcd4b27915782a56d59724d221b0a905c",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..8eae2b2fd2b575494b2a26b35fe8bcdcaa6613cb
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l1_w2_tr.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "accbf3e620ef3341cd2df1f5e28a57440a80966085aa89be4a893c32e40e0537",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:06+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_tr.pkl",
+    "source_sha256": "9e5414c7a03f069ebcedfbaa8f39c20c8f4fe58053edb5f2975a0027d4de60d5",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l2_w2.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee40771b0971e562b04d94f85a8e78e76e1ac190
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_l2_w2.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "92b9aa08bb6c93f31c8f9b22db83502361d4402252ec5d51359bdf6addaec596",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:07+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l2_w2.pkl",
+    "source_sha256": "e2ef5007207ad5e024471b242374cbc8ce049a99cf29d6fe2ae2e443f361df1a",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_tr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..81abcfeb2851fca2b12296d0eaed13fc707d4788
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_meco_tr.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2447ec7bcedf0f6943b832270fc4fbd06b4318974369da155d6a8878ff55a8ba",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:21+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_tr.pkl",
+    "source_sha256": "cac5c2d41d3ae3b91b7d75ab0329c33ea0e4be574beef34c96eafc9b769e5068",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_natural_stories.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..050a0c4a24c07b0e80637f392533f328a24f368c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_natural_stories.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d39173333c51efac8e1e75907cbb7ca747e4ccdd668e6479323c34b0bb935842",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:28+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_natural_stories.pkl",
+    "source_sha256": "c89c74c31cefee0900d11e39ef3a3ce7678ed5140c383759494c7b35090e0993",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_natural_stories_maze.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..5bc47a60cff768c11ac0738d7f233573dee1655e
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_natural_stories_maze.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "39e71d425af1a5e200cb283f659327b85fa2a998fe582e284f5d38774c646340",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:44+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_natural_stories_maze.pkl",
+    "source_sha256": "8ed8b88f3bf83ad3912daede988cbfcbc788c0dc90d6be6d0b9a73eb9137412f",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_onestop.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..a35ee412054ab244e4301933b1fd3512f0241f67
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_onestop.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e7ecb357d51d86e802f87d8dfe0e03784a741ddb9ea7241465965edcff688f5b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:42+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_onestop.pkl",
+    "source_sha256": "8790252cd957e615cadced28c0e4487c41e2495593332ee42ad92e3f9bfb5a1f",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_potec.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..231f78dcbe1281efaeeb8d0c8bbd1a45583ccd69
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_potec.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f38a7359ac7a25f20edc066b44b9e677fec832d5b35a0533fc14a213da2eccd1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:44:51+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_potec.pkl",
+    "source_sha256": "a77be7a9bef7e947d3db707b6e9ff26cf26b78c1c0fa250fcbd5fc0c2ba2e6c4",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_provo.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..a55deaa80024b4afba9ce03f0f49ee61f2234519
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_provo.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "879b990bd8d7c4e427c13a35d768c2fbb6591f1c000bc30f5b3df58d6ac28777",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:19+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_provo.pkl",
+    "source_sha256": "b290f5e906ffe2e8cedfcfa2f7e8fa455de5b2d7ebf726ff2fc87598e431856e",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_sbsat.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..39aafc19d4bd114282bfc7a96c1f4ea5b1daa7f7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_sbsat.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "35b9dedeef3242fcd1e9267c24e50713e79587225df0a77cbb4adb7bb1a01095",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:26:15+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_sbsat.pkl",
+    "source_sha256": "7a725284e1bbf4b6c0bf0ff53ff1dfc1f547a19eb33edc77ad51bd7bd8519a79",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_story_sbsat.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..7946cb8e797aa8745ff807e387e8098877983070
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_story_sbsat.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "70961776d5b7b47495ae89a91991b84ecedcfaa54b9af53bb5e0b42c1ed043a7",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:50+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_story_sbsat.pkl",
+    "source_sha256": "4f05555a37318b22ecf790d0ad9c009a9897c4699c0eff57ea76feeed2c32ae9",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4c515408486f3f5ff01d3d53b8f0a66976d09f6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "875135b0e4a1f7fc45ec22e73ab9f894585df670b0a78a707bcdf440e176ea90",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:52+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_szewczyk_federmeier2022_n400.pkl",
+    "source_sha256": "2e317ece0dc371755ed43a455959486f7bf25ea85e282a9615a705e72d24c702",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_szewczyk_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..b6baaac22e3a5e3b34322f645220e5e933053532
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_szewczyk_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9fbcad63385190404cdc638aa789f32c20e4bffa44ecded0583b052a898b26d6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:24+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_szewczyk_n400.pkl",
+    "source_sha256": "8855b3c746b8cd73cccacc6a1b0ee5c547d2ee7276bac44fc0ff97f4b433d7e4",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_celer.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..950f679786db0b5b30abddbd1d0fac753b01ee87
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_celer.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "837253e56e43f42826fc72235c50fd6824b45935ca39e4fec41f28cef5b6ecbb",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:29:20+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_copco.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1d341602e2698dd87ac4a33a47e0df209577942
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_copco.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8283a225a6d2e61acc23391739e5fb16421010cb30fcf256cb88f9612251b9f0",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_dundee.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..380774b31bb77015a95293c6346fd02a8a2fcb4a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_dundee.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "43bc5d314aa12f860812618dd42f5cfb18e708340dadb197166568219f748af5",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:41+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_federmeier2007_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..de9ea1da78f156db8ecc5f0ac891dd6746a22a67
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_federmeier2007_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "68155f1d7c674f2051b37c699fd25e94864ab1b817822be02577206a295a5af6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:21+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_geco.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..765c4cbf507518cad255662b3236e4ae64150849
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_geco.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "29d7db63d22b94de77859495cf2a12e609b86f056fe4f326e76ebc0933159ad8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:44:41+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_de.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..98efb3d3c251bbbc6e3b865b3d7e18f34197d51f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_de.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "91502de01b59018c27660f925ce36596d7721e6e7eabf8b6802dcd271f6420b9",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:27+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_ee.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..891303370b1963b0c38dfedda4823729bec22e04
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_ee.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "937e42d90b1b395289109d00516700b41560042632e118d8f85d8edb1eae41fc",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:26+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_fi.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0985b636e350df325b50129dd41281a98614d3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_fi.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f1a08b85575feb0d3790f1de25e6e2a61d05cad93969dc9bca3d3d3eac584145",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:27+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_gr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..75f6e690e729e6a00f273f6f652aafb612f41d02
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_gr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e2616e0270593c60866b80e53f827a72ee4e2b879119a69e48d745f6c018f365",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_it.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..86f51566051c54de76cfd9f5934305d67b05198b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_it.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1041f9b8f8a4dab878158b7bd44340b4eda9353309ac4b0db5cfb2165296385c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:32+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_ko.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..1e8120b6719f1c471b6319b46a6b13aa1b0f54e0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_ko.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e113dff38c1921b42126dd5e1d8a9d838b75570848a660bab6a9bf5b48637759",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:34+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ba.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ba.json
new file mode 100644
index 0000000000000000000000000000000000000000..f3882520fb5000a214ebe5553ba57f2efa5bec83
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ba.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "eu",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "659fa9358fbc7b61bc372d3d04318a9dce01b032a697d2251cb7aa241e70e08d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ff25dd55b9d50478ee2845ca3fbb583fca96e6b38b9d89badcf39e12eccb8773",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:22+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_bp.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..c60016099ea15cdd584fc8a6616006f5ddce32e1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_bp.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3b05e04c33a619b72653f4d3f4d820ad26ae79b068e9b03d19d8f6d9d3542949",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:39:51+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ge_po.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..f95b95a71dc3afaace6bccc987d2f9dd2697ed3b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ge_po.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7cfa164b1ed7dd4879f092f386deea81f40d0ce3e83d14f5d9025816629b64a3",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:04+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ge_zu.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..33b0632969e9db5f3a530aae1f6fb175f1882e62
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ge_zu.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "93e3ee6dfc6619b1f49e538b7253307ad00be4eb71ca770b0301ea9368e09ec4",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:07+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_hi_iiith.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_hi_iiith.json
new file mode 100644
index 0000000000000000000000000000000000000000..abef410cc2866b52b84da41f5749d0111de237bf
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_hi_iiith.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "35fb9edae6dbdb289ea9bc78def5ff6fd93a14c5e76ae07c8aa6bb83942954e3",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4b43959b5c1a5260a8213a3b7728e0a180178907106f48dd4e6d722e13871b89",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:15+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_hi_iitk.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_hi_iitk.json
new file mode 100644
index 0000000000000000000000000000000000000000..686f07bd7eb37f34544e2af5dc6291175c03f8b7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_hi_iitk.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "hi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7a9bd43d2aa1205e585c6f89c39f7332f3ba8e3b52b41d6caa8c8153fa376f2d",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "08a78df391c3ce52132337cf1dbb199a372ca591e16295ad02749477e58d55aa",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:16+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ic.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..d9bb3bb056f2251c78620dee06bd72c6ce65425f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_ic.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1403748a80372116299daa4de69ca92bcd4b27915782a56d59724d221b0a905c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_tr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..5174dd7e7f48ceba081fa17d549a9bbc1984b318
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l1_w2_tr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9e5414c7a03f069ebcedfbaa8f39c20c8f4fe58053edb5f2975a0027d4de60d5",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:06+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l2_w2.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0c6f5365d578818537da4b35f6d5a5c67ecfb8f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_l2_w2.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e2ef5007207ad5e024471b242374cbc8ce049a99cf29d6fe2ae2e443f361df1a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:06+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_tr.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..98f80de8c836b55b0d6f4df0afa5277b8e65e767
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_meco_tr.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cac5c2d41d3ae3b91b7d75ab0329c33ea0e4be574beef34c96eafc9b769e5068",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:20+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_natural_stories.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..e13dd1f1c879a856ac132edd81fcc2eeb60c0e07
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_natural_stories.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c89c74c31cefee0900d11e39ef3a3ce7678ed5140c383759494c7b35090e0993",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:28+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_natural_stories_maze.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc1e26bf978c1b4ae2a9597aab162531ebd54d9f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_natural_stories_maze.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8ed8b88f3bf83ad3912daede988cbfcbc788c0dc90d6be6d0b9a73eb9137412f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:43+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_onestop.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..85257bef0eb40f2bb7e2560468c5840b374783b9
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_onestop.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8790252cd957e615cadced28c0e4487c41e2495593332ee42ad92e3f9bfb5a1f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:41+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_potec.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..29aee7ddf01d217142c87cf19a2fe75e292d9c6a
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_potec.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a77be7a9bef7e947d3db707b6e9ff26cf26b78c1c0fa250fcbd5fc0c2ba2e6c4",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:44:46+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_provo.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..9498b66b1e40ab0aca8e1e23a6a5fc6f3fafe206
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_provo.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b290f5e906ffe2e8cedfcfa2f7e8fa455de5b2d7ebf726ff2fc87598e431856e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_sbsat.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..1151cc7773c325f3e8ce3ea8955a930132b309a1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_sbsat.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7a725284e1bbf4b6c0bf0ff53ff1dfc1f547a19eb33edc77ad51bd7bd8519a79",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:26:06+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_story_sbsat.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd13f25e5d827de2e1eeb8bd76362ae4ddec6380
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_story_sbsat.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4f05555a37318b22ecf790d0ad9c009a9897c4699c0eff57ea76feeed2c32ae9",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:49+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..f890aecb8fde147a65ccfa05eb410b84fb56d568
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2e317ece0dc371755ed43a455959486f7bf25ea85e282a9615a705e72d24c702",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:52+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_szewczyk_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..abcd79a696f620f7a0a095d1653fb4cff83bba82
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_szewczyk_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8855b3c746b8cd73cccacc6a1b0ee5c547d2ee7276bac44fc0ff97f4b433d7e4",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:23+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..589a47e5941f5eb684bbb2723627ab59f504b14d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_wlotko_federmeier2012_n400.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ef3d94f8e1ee77387a881e715dd3c7b72ec3fa908ebd9bad4d969f8af1b85c62",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:46+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_zuco1.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..52be0dbff4024da1ec4bdf44dedbf1404a523c6b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_tokens_zuco1.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "11664e373708ab0b4ff759ffd4448e4e09093b105593ff1ddbc221b1be887561",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 512,
+    "stride": 256
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1e867378e6cf97f8baf2b091226f345a434fe1b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_wlotko_federmeier2012_n400.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bf9884ce75b2a3b3279f683716af027567aea02aad92233ba2418415c634d620",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:47+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_wlotko_federmeier2012_n400.pkl",
+    "source_sha256": "ef3d94f8e1ee77387a881e715dd3c7b72ec3fa908ebd9bad4d969f8af1b85c62",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_token_renyi_entropy_a2_zuco1.json b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b04460a649f7438878d139107391bc009e9f6be
--- /dev/null
+++ b/manifests/ai-forever_mGPT_token_renyi_entropy_a2_zuco1.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b7f19fe432f8ce6148ef855f6055ffcb368b6fd0285400899c9466336906399f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:12+00:00"
+  },
+  "derivation": {
+    "source": "ai-forever_mGPT_token_renyi_entropy_a2_tokens_zuco1.pkl",
+    "source_sha256": "11664e373708ab0b4ff759ffd4448e4e09093b105593ff1ddbc221b1be887561",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_celer.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0c6bf32c55f9159645c4808de1a93b63c39f6cd
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_celer.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "9b8841fa4626f3bb01ff1ce806fc9ec5f8ff3ed305fe8653346c88948f105d79",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T04:46:21+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_federmeier2007_n400.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..c445263d4423ca7b20d8c589d99bdb1cef471e45
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_federmeier2007_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "03e36a563cd8ef3cb92685a402f91b286d01e4611fe0deae5674e0536abf9c56",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:07:20+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_ee.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e660ec22d65c62824b3b4ab6106cef09cbfbfd3
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_ee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2b951272ada06bbd0aa554a55f604d6c8d5e49fd0c700401625a33b75149c250",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:09:48+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_fi.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f0d91f3cafdbfbaee6ae40a4935bed353bdd278
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ef37b309e577e36e52eecf0a8ac66eeca4e23231e263785d175f2bc38ff3ca9b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:07:37+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_gr.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..448d449681a1c7258eeeb7e4b84fed58787267db
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_gr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "17cff52d220b92d998c2ede02f7e9cf2378ccbb6e4d0f83659be757023c63fd5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:24:55+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_it.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..9be85332f8593b067bc1a7b8a52b5c6946b3d56d
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "df455835e297818b143f70dc3325d5704f384719041ebb14d1efd76b1bbb3b9a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:23:51+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_tr.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..daa6f1d5117ac433ab88d39af1dea68e745b078b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a4dcdd7e5ed7d3abd5d70df6647b5ae07392e30e7e825547dd503158a4525719",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:59:28+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_natural_stories.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..3fa5827110baab7d5240bbd3278dcdb362f7c053
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_natural_stories.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5b09b5880dae846047a59e13858ccb821973e49a335f4b3304d21363e6b28276",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T20:57:40+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_natural_stories_maze.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ba09f8282d0db48fc3edb974e30897c811ab661
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_natural_stories_maze.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "690000af3ad3a96225400f749184bbbf8ce17bf80d11f863b93a9acd51b24819",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:15:15+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_onestop.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9e662456acc013c5c0d182ae3c24708a1526ac6
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_onestop.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b4a13d7b1c90ecfbe140f5668560bb064ad75298bee80d35906f6eaf8ac9e0b0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:09:22+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_provo.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b77851b067c74e2c2e3dfe00bf817644174e4a0
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_provo.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "9e20ff929eb726bb135d464679d2081f875cdc24981960c17135ddd50f776c2a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:54:15+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_szewczyk_federmeier2022_n400.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ddab064883d78bd39ab8a6ed287a3549b7fd286
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e77bed6183ce01a98a29d045f24d19763ec1de5beeed3f1e452808ff9f41c1b6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:01:17+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_szewczyk_n400.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..762e67d70a53e55687d1a40bad41f6279311a838
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_szewczyk_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c9404f575637d3537215a2200034a69c8db66eee024bf3eca3f19dc6de066cc3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:25:28+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_wlotko_federmeier2012_n400.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..af7bae33bae8bda39cdf38e74fbb56d763546600
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_wlotko_federmeier2012_n400.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4fab4b1d16b34160d82c9ad27392a42ac77e73a52a6245e82d84a1160d86be28",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:51:48+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_zuco1.json b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..972ba1323474ecb79f798ea32964e824ccf15719
--- /dev/null
+++ b/manifests/ai-forever_mGPT_word_infoval_N5_M3_s0_zuco1.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "word_information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 5,
+      "max_tokens": 3,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "500dd0f93721cc2b92cd48731a728858f2462fdc9e930d3cb6cae8e4a51386dd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:04:35+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_ee.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..94145bbaf52f242001497843b266ffae86f3205b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_ee.json
@@ -0,0 +1,68 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal_buggy",
+    "storage_class": "positional",
+    "layer": 2,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "pimentel2024probability"
+  },
+  "status": "completed",
+  "payload_sha256": "f24fd2583f016f3e6fe6b6170f42827d076ad71aef31eeca4e280ebb03130c49",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:03+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_fi.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..68dfa521d61ce50168a854e88ee4105ca9bea7e1
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_fi.json
@@ -0,0 +1,68 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal_buggy",
+    "storage_class": "positional",
+    "layer": 2,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "pimentel2024probability"
+  },
+  "status": "completed",
+  "payload_sha256": "e0e24bd484654d057ffda2cfa782b588b97df7f4d0f322c6f945167c431c4f73",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:23+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_gr.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc73410ba1fd191f35fc12df208fa5c4f0a9a2f7
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_gr.json
@@ -0,0 +1,68 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal_buggy",
+    "storage_class": "positional",
+    "layer": 2,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "pimentel2024probability"
+  },
+  "status": "completed",
+  "payload_sha256": "4b894d51b41f899951c2bb7e0fab2369fb203148e5c0346e9272370e3bb34b3e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:25+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_it.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc0a20db31748cb378fa42b9a306aa2802a981e8
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_it.json
@@ -0,0 +1,68 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal_buggy",
+    "storage_class": "positional",
+    "layer": 2,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "pimentel2024probability"
+  },
+  "status": "completed",
+  "payload_sha256": "6a78daadf3efd78e1d5869034b4c48d29f3fe81928e5497fb8f07a5ca044c055",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:54+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_tr.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..18394f84f7b396bce94b5cd4ed408116eb846065
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_meco_tr.json
@@ -0,0 +1,68 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal_buggy",
+    "storage_class": "positional",
+    "layer": 2,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "pimentel2024probability"
+  },
+  "status": "completed",
+  "payload_sha256": "279bc8a810deed519860372ddc9cf43ec4e7dad316b6975fdc4787556003f2e0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:23+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_potec.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..30848e634783d8b9cd4616086c0c3f4512235e4b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_potec.json
@@ -0,0 +1,68 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal_buggy",
+    "storage_class": "positional",
+    "layer": 2,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "pimentel2024probability"
+  },
+  "status": "completed",
+  "payload_sha256": "aac8a75fa27d57f354e5fc0df9c35bff27af3b04fa82434601c4bc5677b83cf2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:22:59+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_provo.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..07e2cbd3fb36a1d435db1ae4511d8f3101e2c4ea
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_buggy_provo.json
@@ -0,0 +1,68 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal_buggy",
+    "storage_class": "positional",
+    "layer": 2,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "pimentel2024probability"
+  },
+  "status": "completed",
+  "payload_sha256": "23b50ce5675217071db71a17a8af15981a1334286b1b832ce44b4409b50085ae",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:23+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_ee.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..663f0e1bb1eb513b56d988b06a88338ea9dce60b
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_ee.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5240e3ba361bae9b5e9b8c2353d28eaadb2ba9de5303c50d7e2ec485a3f3b11d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:47+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_fi.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..04e0af663198b2b8281f1b0788a8930a9992254f
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_fi.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2cdf3bedab887affb19cd063b3c4d429405f215b92c11e32a5e1ceb8903656dd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:25+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_gr.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..f7d602be0fa93afeca489c7d5d75ff02657ba671
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_gr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b420914355012830762056558661ee7c650735fdf0d2be0266886254abb15f3c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:40+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_it.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..1abfc865b893386c960f6f6593de773875dbb344
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_it.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "58bbd964c0428e845f021746bc228c2e1816fe3ec1f74eea089f44505ff60340",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:59+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_tr.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..81985f4a7b068b7527739eea7d887c4aa38a5cc9
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_meco_tr.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d90829422a98423388de73d4bfaaa18749706878f0a6d97a5cb273e2784040f0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:00+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_potec.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..dbc1c1afd3faf4e0152839c8f9833b06fd83ba39
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_potec.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "613718b471d822b0bccc02c60af147b0d534d8114d20d60f99e51a719751de0a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:43+00:00"
+  }
+}
diff --git a/manifests/ai-forever_mGPT_wordsprobability_surprisal_provo.json b/manifests/ai-forever_mGPT_wordsprobability_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c53b4acb93ac3a56ab5d809fd087fbb93949d1c
--- /dev/null
+++ b/manifests/ai-forever_mGPT_wordsprobability_surprisal_provo.json
@@ -0,0 +1,69 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "ai-forever/mGPT",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "ai-forever/mGPT",
+      "revision": null,
+      "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
+      "files": [
+        "config.json",
+        "merges.txt",
+        "pytorch_model.bin",
+        "special_tokens_map.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "09cec88c14616b4ee584bc00f04861e29dd454890c4f5c5d852738abcf2f2da1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:04+00:00"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_celer.json b/manifests/gpt2-large_logit_lens_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb011a4430b192f80a0a48496e0085ab6ae0c4a8
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_celer.json
@@ -0,0 +1,162 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "05b953843f01628b215f144d07dc5526cb7af8800c3bbad9ce9410a8ef452187",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T22:12:49+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-large_logit_lens_surprisal_tokens_celer.pkl",
+    "source_sha256": "fe23109f2c76261a0363e3dc680981c690044d43dba29eb8219eca7ff66ee4b0",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_dundee.json b/manifests/gpt2-large_logit_lens_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..45df863774e9520639a246094e950579c1497db5
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_dundee.json
@@ -0,0 +1,162 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bc5c21dae58080f6d2bb8e875fa00ff03e83a434c34611338ee29dc05fd5f869",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:18+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-large_logit_lens_surprisal_tokens_dundee.pkl",
+    "source_sha256": "95894292734cead3bd731bc402ad1878eff707120fbaea5d0b0d85f492e1cc15",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_geco.json b/manifests/gpt2-large_logit_lens_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b5773c96406b4a2bad6a3010f4591e5055fa029
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_geco.json
@@ -0,0 +1,162 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "dbc385e8df8223d36e16b7a0f9da67f0c4d84e045982c7d8381c3f6758c7a5bc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:46:48+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-large_logit_lens_surprisal_tokens_geco.pkl",
+    "source_sha256": "430e97e3e4689de7d8eb72c711334361d48686d8f361a42550652eba2c33f5e6",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_onestop.json b/manifests/gpt2-large_logit_lens_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..eefa562bd22b0b1ca7ef0f57782f8287a9d9df7d
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_onestop.json
@@ -0,0 +1,162 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "779165a23b77b2a36edc169f74555668d3ddc37da52f1872b9317a55d8473e46",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:37:43+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-large_logit_lens_surprisal_tokens_onestop.pkl",
+    "source_sha256": "7de7e79042b4f174041969538d2f9ad7ecfc6f10a4bb78dc240ec76ebbd8b172",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_provo.json b/manifests/gpt2-large_logit_lens_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..84894fc55fe6fb8e7a1a1919da758bd11eda3f63
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_provo.json
@@ -0,0 +1,162 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "164f78bf1c08536dd7a9cc9ed2cffa70d68721c1cb88bd4ec122fd585b92b262",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:07+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-large_logit_lens_surprisal_tokens_provo.pkl",
+    "source_sha256": "47531ed3cad613b17fc1827b25456d967c4a120fcb6f202944f4679db17fd2a7",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_sbsat.json b/manifests/gpt2-large_logit_lens_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b266700cc2d6660bd2ff9be66b1eeb244ef13a8
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_sbsat.json
@@ -0,0 +1,162 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "15ecf74216f153555da2a8219e03614f48ac4589486b1f15b73d1114cf778a7c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:58:39+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-large_logit_lens_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "7bab66f0e33b7cf3552bf242358fb6499a4d81123092a63e9c07342ccf3f87d4",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_story_sbsat.json b/manifests/gpt2-large_logit_lens_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..df1fe9a3f7af5f56607d450407758f1f7caa4638
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_story_sbsat.json
@@ -0,0 +1,162 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a26a90c4864e3a68e84b30d6a695dfa29738bc796d7ca8ec0eca8842c6968719",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:28:22+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-large_logit_lens_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "b2ee6d9ee94163288f47267afbc4e4ad5d60604a16b967108aead0965aaea0ea",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_tokens_celer.json b/manifests/gpt2-large_logit_lens_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..99752b319ec5bf4849dc02f293e5843733241866
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_tokens_celer.json
@@ -0,0 +1,121 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fe23109f2c76261a0363e3dc680981c690044d43dba29eb8219eca7ff66ee4b0",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T22:11:41+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_tokens_dundee.json b/manifests/gpt2-large_logit_lens_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1ba4590ee4336cc71fd17b79acb3b77eb6bceb5
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_tokens_dundee.json
@@ -0,0 +1,121 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "95894292734cead3bd731bc402ad1878eff707120fbaea5d0b0d85f492e1cc15",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:00+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_tokens_geco.json b/manifests/gpt2-large_logit_lens_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..641364699cad834fbd028c431df184b1ef2ffa33
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_tokens_geco.json
@@ -0,0 +1,121 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "430e97e3e4689de7d8eb72c711334361d48686d8f361a42550652eba2c33f5e6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:46:24+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_tokens_onestop.json b/manifests/gpt2-large_logit_lens_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..1567114bf5c5fb674b14f1bad926763711f0f2d4
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_tokens_onestop.json
@@ -0,0 +1,121 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7de7e79042b4f174041969538d2f9ad7ecfc6f10a4bb78dc240ec76ebbd8b172",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:37:27+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_tokens_provo.json b/manifests/gpt2-large_logit_lens_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..f563978a557df498810918b53139134cd79ca2ac
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_tokens_provo.json
@@ -0,0 +1,121 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "47531ed3cad613b17fc1827b25456d967c4a120fcb6f202944f4679db17fd2a7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:26:57+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_tokens_sbsat.json b/manifests/gpt2-large_logit_lens_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..4afabf4230147baa556bfd2dd5d94de53ddf9669
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_tokens_sbsat.json
@@ -0,0 +1,121 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7bab66f0e33b7cf3552bf242358fb6499a4d81123092a63e9c07342ccf3f87d4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:58:38+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_tokens_story_sbsat.json b/manifests/gpt2-large_logit_lens_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..688322c2d040b5953855280c2b1272d7d97f5cb7
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,121 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b2ee6d9ee94163288f47267afbc4e4ad5d60604a16b967108aead0965aaea0ea",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:28:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_tokens_zuco1.json b/manifests/gpt2-large_logit_lens_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..145a864f0a84bf171f188420ff9522024a0ffe7a
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_tokens_zuco1.json
@@ -0,0 +1,121 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bd7ebd504babe20d2cc158a7e00d71576d09fa4aa7441472f8599892350f6c8e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:32:48+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-large_logit_lens_surprisal_zuco1.json b/manifests/gpt2-large_logit_lens_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..94ee9cf76b3fee450e2659e3c748676c7bf3f471
--- /dev/null
+++ b/manifests/gpt2-large_logit_lens_surprisal_zuco1.json
@@ -0,0 +1,162 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-large",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-large",
+      "revision": null,
+      "sha256": "34d9071cfbae3054deb008e361ed2c70f8f4a75be31a8775e81b52223b2ba65e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "64f8d561845e627bec07ae0067e22d0a1f95eace1cf3c86312f9f960f7829c55",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:32:58+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-large_logit_lens_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "bd7ebd504babe20d2cc158a7e00d71576d09fa4aa7441472f8599892350f6c8e",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_celer.json b/manifests/gpt2-medium_logit_lens_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..366e0ff43ba6b5414422792222be9b65e3c623ea
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_celer.json
@@ -0,0 +1,138 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "60e3f0350261331ce3ca0bec7d970c6eff93cf8a1546557ffbdfab987d21fd58",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T20:18:56+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-medium_logit_lens_surprisal_tokens_celer.pkl",
+    "source_sha256": "ce39fa32b765e699570eb9001289b9fed547b9fac8edc248c18d74867c2ceaf6",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_dundee.json b/manifests/gpt2-medium_logit_lens_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..a54ff34f98fbd8ba29c35d6d356227f5e2e8555c
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_dundee.json
@@ -0,0 +1,138 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ff8507350f264eed7888cf7283cb65ec3284193be59239b352502cbfb6426a79",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:35+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-medium_logit_lens_surprisal_tokens_dundee.pkl",
+    "source_sha256": "2eef3c980c49c42bf2e5cf465e2e44a60e853c2b90bc20d27043749ba2772121",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_geco.json b/manifests/gpt2-medium_logit_lens_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..b47cd3b95ce8fae72250a74a1567c3d22f2840af
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_geco.json
@@ -0,0 +1,138 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "055f2745818531b53543fb9f418e3c345579752af633ccec164da6008001ef95",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:40:27+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-medium_logit_lens_surprisal_tokens_geco.pkl",
+    "source_sha256": "c0ca4d856908ace35c89791df2b1efac090a904734329af6c798c756e8df0b9c",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_onestop.json b/manifests/gpt2-medium_logit_lens_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..042b7ef33766e61809819653ede9193a649670b0
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_onestop.json
@@ -0,0 +1,138 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "37efd38290aadd26b4a97882cff4674e84eeaf565f52a7e3a95c6a656dd6f67e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:34:11+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-medium_logit_lens_surprisal_tokens_onestop.pkl",
+    "source_sha256": "2fa9544a06a8adc670bd7ad34eac060087ef98998dba0e8e271ec4c3a169090d",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_provo.json b/manifests/gpt2-medium_logit_lens_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..c932193d80174ed2b5cd1b0661998ac7a4751c76
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_provo.json
@@ -0,0 +1,138 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "adac91132fd5a522deb223079067d6cabe2f7aad97d7237b17e97123c78b9ef9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:26:38+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-medium_logit_lens_surprisal_tokens_provo.pkl",
+    "source_sha256": "2b2ce23c5d36813546b61962f040659db77e8cc81e040e43aa60b16d4b858a42",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_sbsat.json b/manifests/gpt2-medium_logit_lens_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0677bcb12d6d4cc6164ff15c5fd9d2a895eb9fc
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_sbsat.json
@@ -0,0 +1,138 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d53a0ff815589412f5b550dba6fccd93a97a2c34c5b7024770b2383851673187",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:58:26+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-medium_logit_lens_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "e721ff17a27d9940ae47ee560f3b7ea8445069b6a2d67fd58ab0ce7fd174f65c",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_story_sbsat.json b/manifests/gpt2-medium_logit_lens_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..c37f5b95a9d37d29c96e763ec0c5d419ee4e1969
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_story_sbsat.json
@@ -0,0 +1,138 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1d212a2d3d3ee78b69740b21bd7f1553734fff6345f923966e2151c047a7c9e6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:16+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-medium_logit_lens_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "0bb2db5fbd934c0e08a5c2eb8d2b79844698e473bc99ffbd11b46787a1940238",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_tokens_celer.json b/manifests/gpt2-medium_logit_lens_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7f9d16cf447cea6fcff76b46e5cf29c6a321e45
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_tokens_celer.json
@@ -0,0 +1,109 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ce39fa32b765e699570eb9001289b9fed547b9fac8edc248c18d74867c2ceaf6",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T20:18:08+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_tokens_dundee.json b/manifests/gpt2-medium_logit_lens_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..55fc5ee14846abf879405f4318c5efc5ea3d12a6
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_tokens_dundee.json
@@ -0,0 +1,109 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2eef3c980c49c42bf2e5cf465e2e44a60e853c2b90bc20d27043749ba2772121",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:19+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_tokens_geco.json b/manifests/gpt2-medium_logit_lens_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..86e2ef527fef9b4a83b375ce7f7b377cd00adf00
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_tokens_geco.json
@@ -0,0 +1,109 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c0ca4d856908ace35c89791df2b1efac090a904734329af6c798c756e8df0b9c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:40:11+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_tokens_onestop.json b/manifests/gpt2-medium_logit_lens_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..fed16ac92155366c1827a4b99b4383d1e48ac757
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_tokens_onestop.json
@@ -0,0 +1,109 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2fa9544a06a8adc670bd7ad34eac060087ef98998dba0e8e271ec4c3a169090d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:33:58+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_tokens_provo.json b/manifests/gpt2-medium_logit_lens_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..97c4d90550f6d040b1f4f713fec4b18edb99a39e
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_tokens_provo.json
@@ -0,0 +1,109 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2b2ce23c5d36813546b61962f040659db77e8cc81e040e43aa60b16d4b858a42",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:26:31+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_tokens_sbsat.json b/manifests/gpt2-medium_logit_lens_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..f063e22a27be47cfc30ab7f6a430230a242f5515
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_tokens_sbsat.json
@@ -0,0 +1,109 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e721ff17a27d9940ae47ee560f3b7ea8445069b6a2d67fd58ab0ce7fd174f65c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:58:25+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_tokens_story_sbsat.json b/manifests/gpt2-medium_logit_lens_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..369fcfb0e24254b08bed0a95f6761d6acd65a6f4
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,109 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0bb2db5fbd934c0e08a5c2eb8d2b79844698e473bc99ffbd11b46787a1940238",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:09+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_tokens_zuco1.json b/manifests/gpt2-medium_logit_lens_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..633d54246097c078591ad4f2d9e373f39e68effe
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_tokens_zuco1.json
@@ -0,0 +1,109 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f5d40dace4d212e5c1da26c2f02101a2f702e1e99c4d7699e26710e86963984c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:30:57+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-medium_logit_lens_surprisal_zuco1.json b/manifests/gpt2-medium_logit_lens_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c13d109579632cc4447ecfffa5dfdcaaa4a18c1
--- /dev/null
+++ b/manifests/gpt2-medium_logit_lens_surprisal_zuco1.json
@@ -0,0 +1,138 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-medium",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-medium",
+      "revision": null,
+      "sha256": "de6cb2234e531156cfb286b1fbd7b4e001800c6a1b863b4c6deb3fb1bde8aa33",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7ec0557a8da34079e3005fcffa171ec149221d80a4639ea83ceb78717514278b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:31:05+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-medium_logit_lens_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "f5d40dace4d212e5c1da26c2f02101a2f702e1e99c4d7699e26710e86963984c",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_celer.json b/manifests/gpt2-xl_logit_lens_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf169065101355ca8b81331fe68ca2934411af54
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_celer.json
@@ -0,0 +1,179 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3a636fae6ba0cee3e53fca14a7305b57099d4ee40e8ead8d14b0cb1ed97a6015",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T00:07:22+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-xl_logit_lens_surprisal_tokens_celer.pkl",
+    "source_sha256": "c213318d3d54c7e17c96d546c6a36ce72e3c83ff6fd89c082fc54141679a7278",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum",
+      "logit_lens_log_p_L37": "sum",
+      "logit_lens_log_p_L38": "sum",
+      "logit_lens_log_p_L39": "sum",
+      "logit_lens_log_p_L40": "sum",
+      "logit_lens_log_p_L41": "sum",
+      "logit_lens_log_p_L42": "sum",
+      "logit_lens_log_p_L43": "sum",
+      "logit_lens_log_p_L44": "sum",
+      "logit_lens_log_p_L45": "sum",
+      "logit_lens_log_p_L46": "sum",
+      "logit_lens_log_p_L47": "sum",
+      "logit_lens_log_p_L48": "sum"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_dundee.json b/manifests/gpt2-xl_logit_lens_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..a63deaf43336dc93463cc369860e00065daad5e6
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_dundee.json
@@ -0,0 +1,179 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1dc4dba21d6f1016eecb3a057bd54931c23cbfcf80ea74db07c36d7db0200775",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:29:24+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-xl_logit_lens_surprisal_tokens_dundee.pkl",
+    "source_sha256": "d2d58019649464c7ce9c57843307d6152c46b73a8929b9b0e4cfb481aac7fe43",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum",
+      "logit_lens_log_p_L37": "sum",
+      "logit_lens_log_p_L38": "sum",
+      "logit_lens_log_p_L39": "sum",
+      "logit_lens_log_p_L40": "sum",
+      "logit_lens_log_p_L41": "sum",
+      "logit_lens_log_p_L42": "sum",
+      "logit_lens_log_p_L43": "sum",
+      "logit_lens_log_p_L44": "sum",
+      "logit_lens_log_p_L45": "sum",
+      "logit_lens_log_p_L46": "sum",
+      "logit_lens_log_p_L47": "sum",
+      "logit_lens_log_p_L48": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_geco.json b/manifests/gpt2-xl_logit_lens_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..f58ddae4c24e16eb75a324cdaeda6ae873412043
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_geco.json
@@ -0,0 +1,179 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "dd3d9ddce3966e8a2c864ef1c9deef11499d8e0b2a1db4d2c2b193b38d6ae5c8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:53:36+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-xl_logit_lens_surprisal_tokens_geco.pkl",
+    "source_sha256": "7fd0a6908f65b26939bc66a20cd5536adf5e2a9d6beee5839fefe5136acd0615",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum",
+      "logit_lens_log_p_L37": "sum",
+      "logit_lens_log_p_L38": "sum",
+      "logit_lens_log_p_L39": "sum",
+      "logit_lens_log_p_L40": "sum",
+      "logit_lens_log_p_L41": "sum",
+      "logit_lens_log_p_L42": "sum",
+      "logit_lens_log_p_L43": "sum",
+      "logit_lens_log_p_L44": "sum",
+      "logit_lens_log_p_L45": "sum",
+      "logit_lens_log_p_L46": "sum",
+      "logit_lens_log_p_L47": "sum",
+      "logit_lens_log_p_L48": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_onestop.json b/manifests/gpt2-xl_logit_lens_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0973e4ce165664a33968c456ddd40bd7844f2bf
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_onestop.json
@@ -0,0 +1,179 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ada4a22d548c168e3da05fd5797958bf9cb061829ec6f3830b17053e6eb7f8db",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:41:17+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-xl_logit_lens_surprisal_tokens_onestop.pkl",
+    "source_sha256": "397685f214b208465ccbb1f87894979306f0d0b7711c0811e40558e26c46c6af",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum",
+      "logit_lens_log_p_L37": "sum",
+      "logit_lens_log_p_L38": "sum",
+      "logit_lens_log_p_L39": "sum",
+      "logit_lens_log_p_L40": "sum",
+      "logit_lens_log_p_L41": "sum",
+      "logit_lens_log_p_L42": "sum",
+      "logit_lens_log_p_L43": "sum",
+      "logit_lens_log_p_L44": "sum",
+      "logit_lens_log_p_L45": "sum",
+      "logit_lens_log_p_L46": "sum",
+      "logit_lens_log_p_L47": "sum",
+      "logit_lens_log_p_L48": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_provo.json b/manifests/gpt2-xl_logit_lens_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..bdd9f35c804cbab6396d0af079a7a7071e9909e2
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_provo.json
@@ -0,0 +1,179 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "454a78e74c7015caac3dc6563f654015c563e4b8ea3f9fae5f76ef72e3a9ae3b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:28:08+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-xl_logit_lens_surprisal_tokens_provo.pkl",
+    "source_sha256": "1ac292291fde5d1d5417a8a1d717fb7c527691426c1bf5c699d6cc42e93d480e",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum",
+      "logit_lens_log_p_L37": "sum",
+      "logit_lens_log_p_L38": "sum",
+      "logit_lens_log_p_L39": "sum",
+      "logit_lens_log_p_L40": "sum",
+      "logit_lens_log_p_L41": "sum",
+      "logit_lens_log_p_L42": "sum",
+      "logit_lens_log_p_L43": "sum",
+      "logit_lens_log_p_L44": "sum",
+      "logit_lens_log_p_L45": "sum",
+      "logit_lens_log_p_L46": "sum",
+      "logit_lens_log_p_L47": "sum",
+      "logit_lens_log_p_L48": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_sbsat.json b/manifests/gpt2-xl_logit_lens_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e4d5643f86aa9b0ed403c19ab98d46d4e726033
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_sbsat.json
@@ -0,0 +1,179 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bdd54045db9746025ed1b38a9e1d53035fb0a8489fffdf103ae0501274c5667d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:59:00+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-xl_logit_lens_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "a533c761f1eb2e3ca6a9c722c147e08c9fff8f6adeb9f6462ff93dff8cb5857f",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum",
+      "logit_lens_log_p_L37": "sum",
+      "logit_lens_log_p_L38": "sum",
+      "logit_lens_log_p_L39": "sum",
+      "logit_lens_log_p_L40": "sum",
+      "logit_lens_log_p_L41": "sum",
+      "logit_lens_log_p_L42": "sum",
+      "logit_lens_log_p_L43": "sum",
+      "logit_lens_log_p_L44": "sum",
+      "logit_lens_log_p_L45": "sum",
+      "logit_lens_log_p_L46": "sum",
+      "logit_lens_log_p_L47": "sum",
+      "logit_lens_log_p_L48": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_story_sbsat.json b/manifests/gpt2-xl_logit_lens_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..e553b89d18bdc7c42f326a4672269d1c5c1b963c
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_story_sbsat.json
@@ -0,0 +1,179 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5848d6c244175655becb59206d550757b4e55cc290e5a70bc77aa47e15383e32",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:28:43+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-xl_logit_lens_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "41effbd41fb5f501c1fe43caca0f32a86408c4e83720b84d96f984c71cdc561b",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum",
+      "logit_lens_log_p_L37": "sum",
+      "logit_lens_log_p_L38": "sum",
+      "logit_lens_log_p_L39": "sum",
+      "logit_lens_log_p_L40": "sum",
+      "logit_lens_log_p_L41": "sum",
+      "logit_lens_log_p_L42": "sum",
+      "logit_lens_log_p_L43": "sum",
+      "logit_lens_log_p_L44": "sum",
+      "logit_lens_log_p_L45": "sum",
+      "logit_lens_log_p_L46": "sum",
+      "logit_lens_log_p_L47": "sum",
+      "logit_lens_log_p_L48": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_tokens_celer.json b/manifests/gpt2-xl_logit_lens_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..65d1cc1a3292f8e95aa6554c59315e4865537832
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_tokens_celer.json
@@ -0,0 +1,126 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c213318d3d54c7e17c96d546c6a36ce72e3c83ff6fd89c082fc54141679a7278",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T00:05:54+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_tokens_dundee.json b/manifests/gpt2-xl_logit_lens_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..bcd4a368cdbf454d8aa20238cc423cf35220134f
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_tokens_dundee.json
@@ -0,0 +1,126 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d2d58019649464c7ce9c57843307d6152c46b73a8929b9b0e4cfb481aac7fe43",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:29:02+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_tokens_geco.json b/manifests/gpt2-xl_logit_lens_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..924235e4dfa6f265b30f767dd418c624212ae870
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_tokens_geco.json
@@ -0,0 +1,126 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7fd0a6908f65b26939bc66a20cd5536adf5e2a9d6beee5839fefe5136acd0615",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:53:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_tokens_onestop.json b/manifests/gpt2-xl_logit_lens_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..394594a09a54729d32d4188e5f789d626c2c6bdc
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_tokens_onestop.json
@@ -0,0 +1,126 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "397685f214b208465ccbb1f87894979306f0d0b7711c0811e40558e26c46c6af",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:40:59+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_tokens_provo.json b/manifests/gpt2-xl_logit_lens_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..0bab2c07233d2b0e47d1be08b5026681201f3eff
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_tokens_provo.json
@@ -0,0 +1,126 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1ac292291fde5d1d5417a8a1d717fb7c527691426c1bf5c699d6cc42e93d480e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:27:57+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_tokens_sbsat.json b/manifests/gpt2-xl_logit_lens_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..de531cf2d1f36650039101107f2624bba8bddf25
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_tokens_sbsat.json
@@ -0,0 +1,126 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a533c761f1eb2e3ca6a9c722c147e08c9fff8f6adeb9f6462ff93dff8cb5857f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:58:59+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_tokens_story_sbsat.json b/manifests/gpt2-xl_logit_lens_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3929e947aa7b1a0cfeeee27e0031efcd8aad322
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,126 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "41effbd41fb5f501c1fe43caca0f32a86408c4e83720b84d96f984c71cdc561b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:28:36+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_tokens_zuco1.json b/manifests/gpt2-xl_logit_lens_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae6b8de1e9f1ed74ad4d963bcf18eb76917f38fe
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_tokens_zuco1.json
@@ -0,0 +1,126 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1af999a03e7aecefd1808c3856adb693c5683ba1528fd910f1dd1094d6431845",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:32:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2-xl_logit_lens_surprisal_zuco1.json b/manifests/gpt2-xl_logit_lens_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4fbd9fa9e09d8363bf773ef67036dbda878c35a
--- /dev/null
+++ b/manifests/gpt2-xl_logit_lens_surprisal_zuco1.json
@@ -0,0 +1,179 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2-xl",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2-xl",
+      "revision": null,
+      "sha256": "1f270a3a87ef792927030a6772e9fb5491ed897747562f08e2d0baa6d020d25e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "generation_config_for_text_generation.json",
+        "merges.txt",
+        "model.safetensors",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats",
+      "logit_lens_log_p_L13": "nats",
+      "logit_lens_log_p_L14": "nats",
+      "logit_lens_log_p_L15": "nats",
+      "logit_lens_log_p_L16": "nats",
+      "logit_lens_log_p_L17": "nats",
+      "logit_lens_log_p_L18": "nats",
+      "logit_lens_log_p_L19": "nats",
+      "logit_lens_log_p_L20": "nats",
+      "logit_lens_log_p_L21": "nats",
+      "logit_lens_log_p_L22": "nats",
+      "logit_lens_log_p_L23": "nats",
+      "logit_lens_log_p_L24": "nats",
+      "logit_lens_log_p_L25": "nats",
+      "logit_lens_log_p_L26": "nats",
+      "logit_lens_log_p_L27": "nats",
+      "logit_lens_log_p_L28": "nats",
+      "logit_lens_log_p_L29": "nats",
+      "logit_lens_log_p_L30": "nats",
+      "logit_lens_log_p_L31": "nats",
+      "logit_lens_log_p_L32": "nats",
+      "logit_lens_log_p_L33": "nats",
+      "logit_lens_log_p_L34": "nats",
+      "logit_lens_log_p_L35": "nats",
+      "logit_lens_log_p_L36": "nats",
+      "logit_lens_log_p_L37": "nats",
+      "logit_lens_log_p_L38": "nats",
+      "logit_lens_log_p_L39": "nats",
+      "logit_lens_log_p_L40": "nats",
+      "logit_lens_log_p_L41": "nats",
+      "logit_lens_log_p_L42": "nats",
+      "logit_lens_log_p_L43": "nats",
+      "logit_lens_log_p_L44": "nats",
+      "logit_lens_log_p_L45": "nats",
+      "logit_lens_log_p_L46": "nats",
+      "logit_lens_log_p_L47": "nats",
+      "logit_lens_log_p_L48": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3d77d8d0a68a180f452272a89c70795955e0e18a14d94bfa7b8c60ead401827a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:32:22+00:00"
+  },
+  "derivation": {
+    "source": "gpt2-xl_logit_lens_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "1af999a03e7aecefd1808c3856adb693c5683ba1528fd910f1dd1094d6431845",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum",
+      "logit_lens_log_p_L13": "sum",
+      "logit_lens_log_p_L14": "sum",
+      "logit_lens_log_p_L15": "sum",
+      "logit_lens_log_p_L16": "sum",
+      "logit_lens_log_p_L17": "sum",
+      "logit_lens_log_p_L18": "sum",
+      "logit_lens_log_p_L19": "sum",
+      "logit_lens_log_p_L20": "sum",
+      "logit_lens_log_p_L21": "sum",
+      "logit_lens_log_p_L22": "sum",
+      "logit_lens_log_p_L23": "sum",
+      "logit_lens_log_p_L24": "sum",
+      "logit_lens_log_p_L25": "sum",
+      "logit_lens_log_p_L26": "sum",
+      "logit_lens_log_p_L27": "sum",
+      "logit_lens_log_p_L28": "sum",
+      "logit_lens_log_p_L29": "sum",
+      "logit_lens_log_p_L30": "sum",
+      "logit_lens_log_p_L31": "sum",
+      "logit_lens_log_p_L32": "sum",
+      "logit_lens_log_p_L33": "sum",
+      "logit_lens_log_p_L34": "sum",
+      "logit_lens_log_p_L35": "sum",
+      "logit_lens_log_p_L36": "sum",
+      "logit_lens_log_p_L37": "sum",
+      "logit_lens_log_p_L38": "sum",
+      "logit_lens_log_p_L39": "sum",
+      "logit_lens_log_p_L40": "sum",
+      "logit_lens_log_p_L41": "sum",
+      "logit_lens_log_p_L42": "sum",
+      "logit_lens_log_p_L43": "sum",
+      "logit_lens_log_p_L44": "sum",
+      "logit_lens_log_p_L45": "sum",
+      "logit_lens_log_p_L46": "sum",
+      "logit_lens_log_p_L47": "sum",
+      "logit_lens_log_p_L48": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_charbeam_surprisal_K5_chars_celer.json b/manifests/gpt2_charbeam_surprisal_K5_chars_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d0e9a6da8cda224361342a78827e37eb2e7d26f
--- /dev/null
+++ b/manifests/gpt2_charbeam_surprisal_K5_chars_celer.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "characters",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "932f02dbd1085e53c1c6e157c24b580882d2b85846f5aed9a495288e0a612ef4",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0f1827232d157f2023459a2d2f4cf9a8a9aa50448fea639ade72ee123a3fab21",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T15:08:10+00:00"
+  }
+}
diff --git a/manifests/gpt2_charbeam_surprisal_K5_chars_geco.json b/manifests/gpt2_charbeam_surprisal_K5_chars_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..865dd35084102a3ab586fe3956c9bcdc94dcd149
--- /dev/null
+++ b/manifests/gpt2_charbeam_surprisal_K5_chars_geco.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "characters",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "82cd6559f6368ab3318188066ca35e82c652f2b3ffeafb8efd4b2066515dbb31",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "76d21a1268bc35d8c5b3a1e47ae3541c1ab08a3e8641933ceab1918c17902a4f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:53:49+00:00"
+  }
+}
diff --git a/manifests/gpt2_charbeam_surprisal_K5_chars_provo.json b/manifests/gpt2_charbeam_surprisal_K5_chars_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..52200750c178be98198f8e53b8531706aa73e3c8
--- /dev/null
+++ b/manifests/gpt2_charbeam_surprisal_K5_chars_provo.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "characters",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "1bfa5187af0e0c6cc0ae7f014b5ab37e291cbfdb6026b082192345e9680692ab",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4804742e7d87a28fcb1e95bcac61f87f3716b2f3a5200934125293cc52b1c1b9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:29:11+00:00"
+  }
+}
diff --git a/manifests/gpt2_charbeam_surprisal_K5_chars_sbsat.json b/manifests/gpt2_charbeam_surprisal_K5_chars_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..402861ff9b79dbf71925c8e558ba4d2f883d9798
--- /dev/null
+++ b/manifests/gpt2_charbeam_surprisal_K5_chars_sbsat.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "characters",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "18381c4a6abea8e2b88683311558217261678ee0596f0858c0064b2ce1ff9433",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "42b6f6829ff3d09d10e09bb794bc63c2747e8a7873a5d1d32b592e40c6489814",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:01:14+00:00"
+  }
+}
diff --git a/manifests/gpt2_charbeam_surprisal_K5_chars_story_sbsat.json b/manifests/gpt2_charbeam_surprisal_K5_chars_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..23c70e7589b70e78d0a30e333783259e8eb6043d
--- /dev/null
+++ b/manifests/gpt2_charbeam_surprisal_K5_chars_story_sbsat.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "characters",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "aa88512bdf880223ee0298262e0bfc0c765346ef5004db5fccd8910a168ba549",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3e178aa880f1a03c3b4ee3d0130c54ecb53d58bc949c7c82642f44e72c8027ba",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:41:19+00:00"
+  }
+}
diff --git a/manifests/gpt2_charbeam_surprisal_K5_chars_zuco1.json b/manifests/gpt2_charbeam_surprisal_K5_chars_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..4980f0b3d367a83d12eec5cdabbd9b7e598c6560
--- /dev/null
+++ b/manifests/gpt2_charbeam_surprisal_K5_chars_zuco1.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "characters",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce258593c1937b299ea92fd744c64e4436fc2b3a6122bfd81766c2f81f9d5fb0",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "65693c4c255bc6c710247c950a302472560b30d784dfa90c3d59e98c6d807241",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:48:07+00:00"
+  }
+}
diff --git a/manifests/gpt2_charbeam_surprisal_K5_geco.json b/manifests/gpt2_charbeam_surprisal_K5_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..d42b152ca1cb0798fcd3065d1c9882e1930b05f5
--- /dev/null
+++ b/manifests/gpt2_charbeam_surprisal_K5_geco.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "12585d0eba189bcc4c807c6817cdf1ad5a26f28db0e05f598ea5faa29056c427",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:53:35+00:00"
+  }
+}
diff --git a/manifests/gpt2_charbeam_surprisal_K5_provo.json b/manifests/gpt2_charbeam_surprisal_K5_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..7097cc3e0a7910b4289656d10fcecd25a257e0b9
--- /dev/null
+++ b/manifests/gpt2_charbeam_surprisal_K5_provo.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "550003a29d3736040f53e8257e9c2fb72fec57fc39231b899c5b388027c41469",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:29:12+00:00"
+  }
+}
diff --git a/manifests/gpt2_charbeam_surprisal_K5_sbsat.json b/manifests/gpt2_charbeam_surprisal_K5_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..fb509aaa5bdae4bc9034229541ce5c9cb43f4e2a
--- /dev/null
+++ b/manifests/gpt2_charbeam_surprisal_K5_sbsat.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9911cdc3b40e8f4f793f5030e049ed52eb7bd1b08f9a8b8c8b4c1dcf47160e56",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:01:32+00:00"
+  }
+}
diff --git a/manifests/gpt2_charbeam_surprisal_K5_story_sbsat.json b/manifests/gpt2_charbeam_surprisal_K5_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..e7327563cb3de314e1450dc080470176d7132985
--- /dev/null
+++ b/manifests/gpt2_charbeam_surprisal_K5_story_sbsat.json
@@ -0,0 +1,75 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "char_beam",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0c4ca2e4cbf526d879c46a200132f43c7793db3762b6f6238ef042155646fdb3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:40:46+00:00"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_celer.json b/manifests/gpt2_cont_entropy_N512_L10_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..9511b6d5fad2060e182255783ce985ce3c5f2b2a
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_celer.json
@@ -0,0 +1,80 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "9a8c9a95b7231e36b6cfb33e3ed9e86c022f56200a900603a3f42f5f27a01b12",
+    "n_stimuli": 28535,
+    "shards": 16
+  },
+  "status": "completed",
+  "payload_sha256": "258a1d52a2a567a668a63333c5a02f56f505bb76de51820d6639c91f9556f0be",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX, Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T08:49:01+00:00"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_dundee.json b/manifests/gpt2_cont_entropy_N512_L10_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ef368fba679820cc83efaae65ba56bc5d094f28
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_dundee.json
@@ -0,0 +1,80 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "55e4d5b6adb8f258a2d2e10aa3b9c8ba3cc43942bba3df2f1fe2ec337793c263",
+    "n_stimuli": 20,
+    "shards": 20
+  },
+  "status": "completed",
+  "payload_sha256": "256cff8a17fd11d55db538bc37482cfdf875c23fde0db525e0704de7ef0ebecc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB, NVIDIA L40S, Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T09:10:49+00:00"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_geco.json b/manifests/gpt2_cont_entropy_N512_L10_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc9c3c3a653021dd7231d2e31e7a5918da8551b2
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_geco.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d0d0418bae7ff5f2306c20c0f2db734d9b4ae8b31fbfe49d09a2074bd4188a44",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T03:31:52+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L10_s0_tokens_geco.pkl",
+    "source_sha256": "6e76683361a32f07bd08903d56a722fb4090c4b7ec38c0fb868570b78e0fe507",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_onestop.json b/manifests/gpt2_cont_entropy_N512_L10_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..503d9e699f90bcde3056a78c9db8b88fcb5715a0
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_onestop.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fe1622c0cebc646b3b15623ebc17e1b87471eaa4a34be5a774a00a55b3b00531",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:48:04+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L10_s0_tokens_onestop.pkl",
+    "source_sha256": "2992da8166e3f9c8474785dc7c3fffd2a9bc97ba384ba9d709765490c733b047",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_provo.json b/manifests/gpt2_cont_entropy_N512_L10_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..cee4c33c5381e2dc6280a79330feca944118529c
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_provo.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6d66dfedc036dd8995f8b2cd374acc22d1c0bb5deea328513bdfa077e3d6ac58",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:41:23+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L10_s0_tokens_provo.pkl",
+    "source_sha256": "893e05be09024da3d2da6b99a1a18007839975fc7a59c2dc515c21582694c36d",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_sbsat.json b/manifests/gpt2_cont_entropy_N512_L10_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..78e685aa420bd59da27484e4db1638614a991932
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "42e28d341f1884b6281f675d15e90b7ccad2dc7cd1c7224e99e2cfeb59ff527b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:19:39+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L10_s0_tokens_sbsat.pkl",
+    "source_sha256": "0600e4bab544f2e5d24cf79e525cc34dfdaa9f6b42aa40655d331bf466627640",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_story_sbsat.json b/manifests/gpt2_cont_entropy_N512_L10_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e6a38f2183489e31962359124429ee0e34033ed
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_story_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d51ea511e82092c37e54a091abde3a55fbb363ce3b5300a06c6b87ce4e4b44c0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:21:03+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L10_s0_tokens_story_sbsat.pkl",
+    "source_sha256": "d2d69280ddb5cf1d75d85018b19cb53fb1efa12c27056514ba8c163706a3867b",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_celer.json b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..273aa7960a736e30b24e9fe35aec89a8b497c35f
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_celer.json
@@ -0,0 +1,80 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "9a8c9a95b7231e36b6cfb33e3ed9e86c022f56200a900603a3f42f5f27a01b12",
+    "n_stimuli": 28535,
+    "shards": 16
+  },
+  "status": "completed",
+  "payload_sha256": "74cb3179f49d373168647eab8639fddee2aa16436650f9ef44b3ec202b4afb92",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX, Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T08:48:59+00:00"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_dundee.json b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..688f2d61445af3d012c9732798f9f387e654c904
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_dundee.json
@@ -0,0 +1,80 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "55e4d5b6adb8f258a2d2e10aa3b9c8ba3cc43942bba3df2f1fe2ec337793c263",
+    "n_stimuli": 20,
+    "shards": 20
+  },
+  "status": "completed",
+  "payload_sha256": "8c755812518cfd5223d7399a21edd0915c4072185c43505787f8b158087d2f22",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB, NVIDIA L40S, Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T09:10:48+00:00"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_geco.json b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..313bdb6620128ea3502b64d95097dcba78301117
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_geco.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6e76683361a32f07bd08903d56a722fb4090c4b7ec38c0fb868570b78e0fe507",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T03:31:46+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 502,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_onestop.json b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..a19f8344e6d2de3049e7d3332fbf7804d781b13c
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_onestop.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2992da8166e3f9c8474785dc7c3fffd2a9bc97ba384ba9d709765490c733b047",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:48:00+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 502,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_provo.json b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..eead1f2debb333f9634da3d022a2dbc966e45f0c
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_provo.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "893e05be09024da3d2da6b99a1a18007839975fc7a59c2dc515c21582694c36d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:41:22+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 502,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_sbsat.json b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..3651579c77517996fe520dc024eceeb07f706dff
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_sbsat.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0600e4bab544f2e5d24cf79e525cc34dfdaa9f6b42aa40655d331bf466627640",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:19:38+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 502,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_story_sbsat.json b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b5cbaae7cb7fdcd3f7dcf5e4c8467792d08b265
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L10_s0_tokens_story_sbsat.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d2d69280ddb5cf1d75d85018b19cb53fb1efa12c27056514ba8c163706a3867b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:21:02+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 502,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_celer.json b/manifests/gpt2_cont_entropy_N512_L5_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..9484a2975f6b2916f0892a878452028be2ec4fe8
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_celer.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "583514c7ca433c0f4dc1718b99f3c9e2e5ef783d26b89100e0ff28b46ed3e287",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T07:10:20+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_celer.pkl",
+    "source_sha256": "ff307131e3a58d3330b20fbec4bd8d889ae50f7e896e7f444e01745976c67378",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_dundee.json b/manifests/gpt2_cont_entropy_N512_L5_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..4360a06bf1e8fd460c7d8f8f74297c9cba4f705c
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_dundee.json
@@ -0,0 +1,80 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "55e4d5b6adb8f258a2d2e10aa3b9c8ba3cc43942bba3df2f1fe2ec337793c263",
+    "n_stimuli": 20,
+    "shards": 20
+  },
+  "status": "completed",
+  "payload_sha256": "c4c8f7c0c2cc1dd13f0ac154c2816325903a97d5ee71ddc5775b995e9aceb6d7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-40GB, NVIDIA L40S, Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T09:31:06+00:00"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_federmeier2007_n400.json b/manifests/gpt2_cont_entropy_N512_L5_s0_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..e89ea2ee903609ac281df360efbf1a6c86e23523
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_federmeier2007_n400.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2281bb9ed9eb4f9562b82c46ee55e4157a3f2e12474a37af06f5d13573289608",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:36:33+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.pkl",
+    "source_sha256": "2befe502a2bf88402f182893636687a522d4963aac191810c6fe85a1623cedfa",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_geco.json b/manifests/gpt2_cont_entropy_N512_L5_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e4326322aff8b3cd51a5fa01cebce6bf0446b83
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_geco.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f42a9972103b382de161307ad6bb46082bbe40c9ef39624a4107328d0c71c2a9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T21:01:42+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_geco.pkl",
+    "source_sha256": "9cdf52b92ffbe4bd005d681ef8d76d01886af0a987b8181686152c92d3c69e51",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_meco_l2_w2.json b/manifests/gpt2_cont_entropy_N512_L5_s0_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8658bc1d8967d6155486a10bae3cff7a9afbd9d
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_meco_l2_w2.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2dade104df1544221552e4d454c9219c428dbf6bfaea04389c23a686b268dea2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:02:00+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.pkl",
+    "source_sha256": "c504804d6a05504b889c7410ca44412c87806af38761cc8068c5a9bb6bb342e3",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_natural_stories_maze.json b/manifests/gpt2_cont_entropy_N512_L5_s0_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a4e0551eeef19d399d027faaa3ac62e9f8722d9
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_natural_stories_maze.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c5cd87d801c94b2a61efd607389113894058994665ad998465fc50f68925697f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:24:26+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_natural_stories_maze.pkl",
+    "source_sha256": "59465977f0a800c1ca5fc2e8b8018f8866c40febef08d869d3d4239789cc0837",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_onestop.json b/manifests/gpt2_cont_entropy_N512_L5_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..57753cf605d5164836f7eac839a17ad6b9e21498
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_onestop.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3d309ed57ed2e8c17ac098c2d0261b1b5b57f0a18c474c7e1d3ad674aa9a35a7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:01:13+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_onestop.pkl",
+    "source_sha256": "e74094b2cbe644f17f93f0c5a9f4231fad9fdadec10b6a4aa969cb23568f4616",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_provo.json b/manifests/gpt2_cont_entropy_N512_L5_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..900c537071fb625a3cd9a2eba8eaf40b330fc802
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_provo.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c6630fd9b040a1d8bb5272253a7062c0a83a7e7a1b613afa81d501049e1090be",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:26:06+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_provo.pkl",
+    "source_sha256": "522f9198fbce3db7bf0fd893a01450419db078b69c9b63a8b13008cccf01b46e",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_sbsat.json b/manifests/gpt2_cont_entropy_N512_L5_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..feda04d0659e2bc2eec4a9e8265be62cd00fbf77
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "16ed32d39806b372afa8042520b06ae4d01b9790ed3c3fbe80b918b7081f82d6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:07:41+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_sbsat.pkl",
+    "source_sha256": "51c5b886e95854ccecadb5219ae92482cfdb43a10c585914cb63dfb1146e6a14",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_story_sbsat.json b/manifests/gpt2_cont_entropy_N512_L5_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..e1132d4f811a1c8ae8a753481af00b1f1cf75ada
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_story_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c0b27bbac6bce6700c6d02b4575ab782575f28d032a8150b50bf8c7716780313",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:23:52+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_story_sbsat.pkl",
+    "source_sha256": "ae010d2f01d633522c45e667d416378c1eb36712affaac9c055ceb19de2ddc5c",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_szewczyk_federmeier2022_n400.json b/manifests/gpt2_cont_entropy_N512_L5_s0_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..29669455b9db4c86ca8839065843d86adb3a58f0
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e26e04f4bf214c023a26c4c339fa8635e6f0570178a0bb5124fe30e31be27255",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:00:02+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_federmeier2022_n400.pkl",
+    "source_sha256": "f8f590601a427bfc31222c2b9ec1b93e0cca928e8807ea9e1761013724bede5f",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_szewczyk_n400.json b/manifests/gpt2_cont_entropy_N512_L5_s0_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..049795f95d96e41b207126b5695ff82747d30b78
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_szewczyk_n400.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c46ac468fa8c1da2bcbde69ab38b2cc08031de8dbf329b61e828fda7c4126695",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:26:51+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_n400.pkl",
+    "source_sha256": "427ca3ea017a4bf37dee3c451ad078aa2041d402afad1762e88af4a91465687e",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_celer.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..d06545ce57ec000d17b3134d950042eae48c8aec
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_celer.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ff307131e3a58d3330b20fbec4bd8d889ae50f7e896e7f444e01745976c67378",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T07:09:54+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_dundee.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5b34d1dbcc182807da0e791cc8fd2f10d503609
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_dundee.json
@@ -0,0 +1,80 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    },
+    "corpus_sha256": "55e4d5b6adb8f258a2d2e10aa3b9c8ba3cc43942bba3df2f1fe2ec337793c263",
+    "n_stimuli": 20,
+    "shards": 20
+  },
+  "status": "completed",
+  "payload_sha256": "0e48cb6518228b510ca9891ad7d9110e1df3ab364631be547a6c4cd1c648d661",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-40GB, NVIDIA L40S, Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T09:31:04+00:00"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..10809a838821b294ab623f9da2cd85f611dd896c
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_federmeier2007_n400.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2befe502a2bf88402f182893636687a522d4963aac191810c6fe85a1623cedfa",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:36:23+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_geco.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..9479dd12efa58644aed24cd34f6d99cc376a211f
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_geco.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9cdf52b92ffbe4bd005d681ef8d76d01886af0a987b8181686152c92d3c69e51",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T21:01:29+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b121e460727bcaae207da17e86ee832c2431f17
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_meco_l2_w2.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c504804d6a05504b889c7410ca44412c87806af38761cc8068c5a9bb6bb342e3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:01:58+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_natural_stories_maze.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..d449a7efe93d17b88f07389994e48e8eb3c9956f
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_natural_stories_maze.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "59465977f0a800c1ca5fc2e8b8018f8866c40febef08d869d3d4239789cc0837",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:24:25+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_onestop.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea6528666fcdf9aa3229078116f12347ff33552c
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_onestop.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e74094b2cbe644f17f93f0c5a9f4231fad9fdadec10b6a4aa969cb23568f4616",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:01:10+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_provo.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..9920db904a55ede18e15d16e0c4e778fcfd6e135
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_provo.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "522f9198fbce3db7bf0fd893a01450419db078b69c9b63a8b13008cccf01b46e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:26:05+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_sbsat.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..6405d514d33bd4e048a1b0da1a7135efbc711a8f
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_sbsat.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "51c5b886e95854ccecadb5219ae92482cfdb43a10c585914cb63dfb1146e6a14",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:07:40+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_story_sbsat.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..3fb494c8a42e3629b1dc5d1edfd2a9c8305d9254
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_story_sbsat.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ae010d2f01d633522c45e667d416378c1eb36712affaac9c055ceb19de2ddc5c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:23:49+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_federmeier2022_n400.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a952ddab59c9ec2438e08114f8ddfb33b75832d
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f8f590601a427bfc31222c2b9ec1b93e0cca928e8807ea9e1761013724bede5f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:59:52+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_n400.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..fbb4cea2282ec99485f25d7a0765df404566a822
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_szewczyk_n400.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "427ca3ea017a4bf37dee3c451ad078aa2041d402afad1762e88af4a91465687e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:26:41+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..68974f48b42ee585f0955f88f791e3c78258fdc7
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1f83139025155cb61aabc392e9de4ab2403a41acf6caf1ee197d7c7aa2c78568",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:52:42+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_zuco1.json b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..58077a3cc5fe15eedcfd7ff4593620cd887633b7
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_tokens_zuco1.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b067cf2533e9bcfac104ec215881121eeeb509a4bb0278a2e7f03bcc2471dd92",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:19:52+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 507,
+    "continuation": 5
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.json b/manifests/gpt2_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..026bfa9a3b52b54e30fab06a0c4f3139b70a35da
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_wlotko_federmeier2012_n400.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "111d680325fd0002fb39330de76387f10648479ae2198a0806e23b3aa09b55ac",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:52:52+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_wlotko_federmeier2012_n400.pkl",
+    "source_sha256": "1f83139025155cb61aabc392e9de4ab2403a41acf6caf1ee197d7c7aa2c78568",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N512_L5_s0_zuco1.json b/manifests/gpt2_cont_entropy_N512_L5_s0_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..149cdfc7697b14948a94c4a68214e9f2380f146a
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N512_L5_s0_zuco1.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 5,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fec0bc099b66d68a59bdd57febece6dd47330d89131a9af9bc8ce40f71aa2884",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:19:54+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N512_L5_s0_tokens_zuco1.pkl",
+    "source_sha256": "b067cf2533e9bcfac104ec215881121eeeb509a4bb0278a2e7f03bcc2471dd92",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_celer.json b/manifests/gpt2_cont_entropy_N64_L10_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..32ce7cda29c77c92cb69cb645b0f0bb6c86f7bcb
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_celer.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "237c6a1ddde9a67497ac0d8937dad484850b1758e77883d5a7fecdb25a7ffc5d",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T01:02:41+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N64_L10_s0_tokens_celer.pkl",
+    "source_sha256": "dbf4d5adcf7c10515caa9c934055f5810b7571368481c89196a25df0d0a687ea",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_dundee.json b/manifests/gpt2_cont_entropy_N64_L10_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..e23bf6216ad1a5f35030745d8f0300163a39bcf4
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_dundee.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "312d5646da7c474bb9fea91ab7f1d8b87c03cb67c956bbfbc3d7b40465b77491",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:46:53+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N64_L10_s0_tokens_dundee.pkl",
+    "source_sha256": "15eb5fd683c8ced9744e4c6480aa1758b71901f7576191607405e6447cb4a9d1",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_geco.json b/manifests/gpt2_cont_entropy_N64_L10_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..87ad97ddd1c8fedb22dd15b27d239fff347817e5
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_geco.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "034cd081cf2759ae0308ef88ab884e501de26cba67fc86969f85bf4ae1254cab",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:29:39+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N64_L10_s0_tokens_geco.pkl",
+    "source_sha256": "75f225579263d6acb5f4dbb628bbb3d74bddbe6513680bf9914b59ef37a23bbb",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_provo.json b/manifests/gpt2_cont_entropy_N64_L10_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..6cbaa24f8d3e7dd27eba08e743a5721ac343f59b
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_provo.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ccd469c1bf6ceec50810c3085b7dcdc60e3335dd421be6141ec98c1624ca0966",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:28:23+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N64_L10_s0_tokens_provo.pkl",
+    "source_sha256": "8ef11e61f93e66193c61ef8b50b6d2029053c4b9e9cb6fa1eceabcf906bac246",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_sbsat.json b/manifests/gpt2_cont_entropy_N64_L10_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1201400b9405ff2e34f45a4b474aa19931b8deb
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a4b412df377c3765b9de9dc4b3d6d882be2f06af85c513b81997550824a8701a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:27+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N64_L10_s0_tokens_sbsat.pkl",
+    "source_sha256": "5b6cbb6a73cecb2eb82c14c4afa45e73da4722dd6346b0d60d2d35b368fa8336",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_story_sbsat.json b/manifests/gpt2_cont_entropy_N64_L10_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..aad35e45fdefbbe8b2961023014790a5fba911d4
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_story_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "143268dc873cf9d2ff241017a3d50589ba3c6c6db77c6e0805da5a919ac7f499",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:12:38+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_cont_entropy_N64_L10_s0_tokens_story_sbsat.pkl",
+    "source_sha256": "022aba265d3637a11ae43592ae54e3963953506d98eeb4e3d328edd1d55f76bf",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "continuation_entropy": "first",
+      "log_p_samples_nats": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_celer.json b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..f50c73b99fc3dd09fee43450ac486204b4f8e73d
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_celer.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "dbf4d5adcf7c10515caa9c934055f5810b7571368481c89196a25df0d0a687ea",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T01:02:25+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 502,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": 64
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_dundee.json b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..299cd70bd6823633594da76698b0db62e21d0bce
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_dundee.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "15eb5fd683c8ced9744e4c6480aa1758b71901f7576191607405e6447cb4a9d1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:46:50+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 502,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_geco.json b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..356e9e347e1ef1fae0e4f26ccd33716eb8f74f43
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_geco.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "75f225579263d6acb5f4dbb628bbb3d74bddbe6513680bf9914b59ef37a23bbb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:29:27+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 502,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_provo.json b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..7858b99fb21d03cc03059cc6d4148a05f25350c0
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_provo.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8ef11e61f93e66193c61ef8b50b6d2029053c4b9e9cb6fa1eceabcf906bac246",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:28:15+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 502,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_sbsat.json b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..30687d5a9d78eb281c21438d5eebafa3001d2d1d
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_sbsat.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5b6cbb6a73cecb2eb82c14c4afa45e73da4722dd6346b0d60d2d35b368fa8336",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:57:26+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 502,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_story_sbsat.json b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..47a6f2d5d6982709ca074db6d060e6b92f77a191
--- /dev/null
+++ b/manifests/gpt2_cont_entropy_N64_L10_s0_tokens_story_sbsat.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "continuation_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "continuation_entropy": "bits",
+      "log_p_samples_nats": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "022aba265d3637a11ae43592ae54e3963953506d98eeb4e3d328edd1d55f76bf",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:12:29+00:00"
+  },
+  "window": {
+    "rule": "half_windows",
+    "max_positions": 1024,
+    "half_size": 502,
+    "continuation": 10
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "per stimulus from the seed, the model digest and the text; uniforms keyed by token position and sample index",
+    "scheme": "model-bundle-v1",
+    "chunk_size": null
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_celer.json b/manifests/gpt2_ft_entropy_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..b446a90909874b74430735926a18ceda6c85673d
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_celer.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "14b5dfd79db193318d13d65bebc6fa38e2faa5e7b4fff915ed8b967a99b69ed2",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:03:44+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_celer.pkl",
+    "source_sha256": "f2839153ad92c7ee2ff6c9445d26d015762e29890c550fffa52dbd445aff9dcc",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_dundee.json b/manifests/gpt2_ft_entropy_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..443a343f7ec8727b9f3ccf4c694eb34e636aeca3
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_dundee.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "825dff58f20f6e8303a0d96a18bb834d7a35f844d82e738c31433c144038d0a8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:42+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_dundee.pkl",
+    "source_sha256": "4657473c7a46562589e4f2b479f13efbfe088296750d86e22714cb5b9b4083f2",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_federmeier2007_n400.json b/manifests/gpt2_ft_entropy_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..868e0b06529c1d8a7cdb2b4a6c6de8adee227b52
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_federmeier2007_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5af033941f7a497471378702a1e0340964e999ab7ed54158799644c15547ec2d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:47+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_federmeier2007_n400.pkl",
+    "source_sha256": "5995de923a73e13502cabe62caf718c77d3c325127dd6db81a7b57ac0ac60b19",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_geco.json b/manifests/gpt2_ft_entropy_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..942375618b84448364afae2af8dfe9b1a5d5b144
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_geco.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8aa558c32dfa197d5f8ad84112bbc09ace42891b542a1f361cc894ff9dc0c300",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:45:33+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_geco.pkl",
+    "source_sha256": "44e7ac88f89fad28ae0f3accf53b63fac9648acf2bad510eebe7f93e11bfb119",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_meco_l2_w2.json b/manifests/gpt2_ft_entropy_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..5166ec464e14e532f41c5ebae71f96e4e0b021d5
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_meco_l2_w2.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "138d152c9e1fa404568beac005162616f94f1fac31803f075782f27db1b7bf45",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:42+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_meco_l2_w2.pkl",
+    "source_sha256": "3f8c1a651b957468842f7583123aca7adfe7c5e0ffb9554f48561f699be36119",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_natural_stories.json b/manifests/gpt2_ft_entropy_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc492b20de7a5de02d28fbe53420a0b85a743816
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_natural_stories.json
@@ -0,0 +1,74 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": true,
+      "model_identity": null
+    },
+    "model": {
+      "source": "openai-community/gpt2",
+      "revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e",
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ]
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "ca57e94049031d6937ec92f01c668130b26b03ef357c33b5b753f91a783d6626",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3975c851d8c7b8c3f95f8e0e1759a2310962b9d042f4f7a1dd196a94393e5152",
+  "environment": {
+    "code_commit": "dab4617575e68ca1825e76854a713f747f1900bb",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-23T14:18:05+00:00"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_natural_stories_maze.json b/manifests/gpt2_ft_entropy_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..c18ef0d6a69c6ea8e8ed4fe5c96e0ae978ffcd36
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_natural_stories_maze.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4e0891872e05ff3e29a8e732ba76c253102c2ca4f715a47a50342fe116bb6e54",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:53+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_natural_stories_maze.pkl",
+    "source_sha256": "7879a7e60424846cc7a6a5d9b8611eba66df362dba77b5e04eef8ae67bbbbc9f",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_onestop.json b/manifests/gpt2_ft_entropy_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..d42b67c98b1d69c9d217c766dbd937d699c1e16d
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_onestop.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c4a824acb9a55103cddff5c80fa7e0e269f9d90e6ec4ed739ae0300346c6a9ed",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:32+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_onestop.pkl",
+    "source_sha256": "801fb8bff8fcf6af2ebafa5e3c7e4aeccec982164250436abd21cc96859f2d40",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_provo.json b/manifests/gpt2_ft_entropy_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..baef894bbcb3f24fb6c6e5e23d0aae3081e3bd9e
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_provo.json
@@ -0,0 +1,73 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": true,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "60a20ddf8c3d6abf0a88cc1c7f77232125591e734fbe9bb6f29823e3c1f987f3",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:54:00+00:00"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_sbsat.json b/manifests/gpt2_ft_entropy_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..a71de85db66f3d7442563e660177f82e32b8ecbd
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0f338836e083168ff6a2fdf1a654e53c9ff8e4446b90ea1d3cb700f822904c0c",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:26:11+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_sbsat.pkl",
+    "source_sha256": "16b9b5f9ce132cac7da95f301af0dfd5ec18f06fe491bcd737efb67bd3fcf495",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_story_sbsat.json b/manifests/gpt2_ft_entropy_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..87dd25e56f49f14fb2a97593e2bffbce750291e6
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_story_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f4ab1cafadd937f0e20e0d5bf7316831f722687eb880b6273343241c0e02394b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:37+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_story_sbsat.pkl",
+    "source_sha256": "e1e193fab8d96bef5db73883c829a6a7614f8c5a1b2c10534423a9964bf08287",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_szewczyk_federmeier2022_n400.json b/manifests/gpt2_ft_entropy_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..a174bbf2df77182ea2127316a366a3aceff4fa27
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "586608f611f8fd195eea0d1728b5b183ef34391b0fd5c234a6ee7fc206b8f58d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:52+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_szewczyk_federmeier2022_n400.pkl",
+    "source_sha256": "cd712b97e34321b0b39f7c4b11e45a2c864239991521be2f7a8e4a513076d282",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_szewczyk_n400.json b/manifests/gpt2_ft_entropy_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc79b069cb0e261c42b9ae6be08b5e46680fd476
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_szewczyk_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "25903606757c0840f338817062b9f8b19c0ab943d410cfb39a6f3210d23232a2",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:32+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_szewczyk_n400.pkl",
+    "source_sha256": "facd10235e276af5e5ada3a45261b2f06cdc794c88ac02bfbbb1ea44f0f72898",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_celer.json b/manifests/gpt2_ft_entropy_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..95685efb1b0ab0a691663e78c53c27798067a17f
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_celer.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f2839153ad92c7ee2ff6c9445d26d015762e29890c550fffa52dbd445aff9dcc",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:03:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_dundee.json b/manifests/gpt2_ft_entropy_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..9ca08c51dc59ddadb8d28beaa923921b0252ea4c
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_dundee.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4657473c7a46562589e4f2b479f13efbfe088296750d86e22714cb5b9b4083f2",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:40+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_federmeier2007_n400.json b/manifests/gpt2_ft_entropy_tokens_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..e783e98833ec0a144a1c2da7fa81b0922f4aee38
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_federmeier2007_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5995de923a73e13502cabe62caf718c77d3c325127dd6db81a7b57ac0ac60b19",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:46+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_geco.json b/manifests/gpt2_ft_entropy_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..f294a210fbb33cc3d8f7fae28f910fd11d056413
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_geco.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "44e7ac88f89fad28ae0f3accf53b63fac9648acf2bad510eebe7f93e11bfb119",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:45:32+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_meco_l2_w2.json b/manifests/gpt2_ft_entropy_tokens_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..1798d0c708355cc5817601abdb480c30b01d8c43
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_meco_l2_w2.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3f8c1a651b957468842f7583123aca7adfe7c5e0ffb9554f48561f699be36119",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:42+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_natural_stories.json b/manifests/gpt2_ft_entropy_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..2d23be4f6aed60077dede4339353d72014efb78b
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_natural_stories.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9ddca3381918f3bad1ed66bfcbfb31685fad21f3fba4337ba51cb7008511c261",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:40+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_natural_stories_maze.json b/manifests/gpt2_ft_entropy_tokens_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..469e6189e95bb0c52551ed3c593b044719351de5
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_natural_stories_maze.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7879a7e60424846cc7a6a5d9b8611eba66df362dba77b5e04eef8ae67bbbbc9f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:52+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_onestop.json b/manifests/gpt2_ft_entropy_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d6a9e45e949936794cb1352f7f004e1d25d3160
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_onestop.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "801fb8bff8fcf6af2ebafa5e3c7e4aeccec982164250436abd21cc96859f2d40",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:31+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_provo.json b/manifests/gpt2_ft_entropy_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..dce07c127c681d9b6172be2734e024d9931b18e3
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_provo.json
@@ -0,0 +1,82 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "openai-community/gpt2",
+      "revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e",
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ]
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "ca57e94049031d6937ec92f01c668130b26b03ef357c33b5b753f91a783d6626",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "530765ef8f8d695ca02139a579036ee87dd0be041fe48f76300be1b2deb16251",
+  "environment": {
+    "code_commit": "dab4617575e68ca1825e76854a713f747f1900bb",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-23T13:37:32+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_sbsat.json b/manifests/gpt2_ft_entropy_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..003747a7a11bd82555faabcf27d1e27b7d0765b3
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_sbsat.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "16b9b5f9ce132cac7da95f301af0dfd5ec18f06fe491bcd737efb67bd3fcf495",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:26:02+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_story_sbsat.json b/manifests/gpt2_ft_entropy_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..68f390ab07fbc532203f6d216288d108500c51d5
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_story_sbsat.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e1e193fab8d96bef5db73883c829a6a7614f8c5a1b2c10534423a9964bf08287",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_szewczyk_federmeier2022_n400.json b/manifests/gpt2_ft_entropy_tokens_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b5ea5d3194612bd99e8a8b73c034a238f1d7a6f
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cd712b97e34321b0b39f7c4b11e45a2c864239991521be2f7a8e4a513076d282",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:44+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_szewczyk_n400.json b/manifests/gpt2_ft_entropy_tokens_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..cc571d51a2f26ff1d5a7543a291795b9b91d0aa5
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_szewczyk_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "facd10235e276af5e5ada3a45261b2f06cdc794c88ac02bfbbb1ea44f0f72898",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:24+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_wlotko_federmeier2012_n400.json b/manifests/gpt2_ft_entropy_tokens_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..dbdd8bda7a6a44864cf2af05480a1d16527078e5
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_wlotko_federmeier2012_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a353d484502473803bd86e3d7bcf7b8ed29e29721a9b53f9d0e0274f6aa85185",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:57+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_tokens_zuco1.json b/manifests/gpt2_ft_entropy_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..275bb517650ef3ea83c60a16b34ed4abee792d06
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_tokens_zuco1.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ef310984d1d930702f3b822b9f5d536dceaaf2751611ea55e95c9b41128a4b70",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:44:57+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_wlotko_federmeier2012_n400.json b/manifests/gpt2_ft_entropy_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..58dcda9857dd3cbe62671123b497ca4bb4acf96b
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_wlotko_federmeier2012_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cae28c5075d7d17587f600fe3c38a95362e64d2a21264b82d04855a3a5e9adbe",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:40:57+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_wlotko_federmeier2012_n400.pkl",
+    "source_sha256": "a353d484502473803bd86e3d7bcf7b8ed29e29721a9b53f9d0e0274f6aa85185",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_ft_entropy_zuco1.json b/manifests/gpt2_ft_entropy_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..1353ea794bd087a6f0a5c32cf19afffc9f70751a
--- /dev/null
+++ b/manifests/gpt2_ft_entropy_zuco1.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "first_token_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_entropy": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7fe45d02d35df9e65966743b27b7e63856d82a981776fabfd1634e9b99f17bf3",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:45:03+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_ft_entropy_tokens_zuco1.pkl",
+    "source_sha256": "ef310984d1d930702f3b822b9f5d536dceaaf2751611ea55e95c9b41128a4b70",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_celer.json b/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..c064d2d1b46a79181587d57da80645eb200f053e
--- /dev/null
+++ b/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_celer.json
@@ -0,0 +1,78 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words",
+    "corpus_sha256": "9a8c9a95b7231e36b6cfb33e3ed9e86c022f56200a900603a3f42f5f27a01b12",
+    "n_stimuli": 28535,
+    "shards": 16
+  },
+  "status": "completed",
+  "payload_sha256": "3b3e3f4d0d24c4ad3ea964207ea5bc1ebab9b71946145457cd9384b3d428283c",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX, Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T19:02:23+00:00"
+  }
+}
diff --git a/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.json b/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..e298769734d2eb103e6b02199c0e66e172989c06
--- /dev/null
+++ b/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_provo.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a9209fe346d4c936b4375aa2e3866a4d343e1fd9b017cd43d95aafe6f5c31172",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:28:21+00:00"
+  }
+}
diff --git a/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.json b/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..53bfeaf7e5d48ef7a9e3d1f04717b8b64fec18f0
--- /dev/null
+++ b/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a86eb15fd24f06fd474417a48e82fd1189cfab7f1c667fc9cc578e9dec9b5780",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:50:23+00:00"
+  }
+}
diff --git a/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.json b/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..495cd8d4dd0dbdd6d25236c21685b713f93b29c7
--- /dev/null
+++ b/manifests/gpt2_fwd_word_lookahead_entropy_N512_M30_K5_s0_story_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "forward_word_lookahead_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 30,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a0ad5976b01c4a9ef3b28b7a7c6d9b00b1e103a7f864930bafda36f07285f4e1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:02:10+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_dundee.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c0d30288ec5809edd2c494bfe0bbea0c9b0f99d
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_dundee.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "ecea4dfe3847e2b3ddddfaef7ef071619afed6d524c6ec535b2c3f4a42b66ead",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:21:45+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_federmeier2007_n400.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..865a530838735f21a75536f2140802f763c90833
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_federmeier2007_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "ff25ea0dc2d4657a28a0009fd5b90f8dfdc41ed568984ec84be639f46778f8b2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:23:35+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_geco.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..550b63038a671cc34c0232e3ac00b0dd485acfcf
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_geco.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "00f1e5ee02a8c6129eb28e964234ee60a267d69e196c176809a075df00f5b252",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:44:37+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_meco_l2_w2.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..e507230ea51269ea10ae7219c0584d714ab04bac
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_meco_l2_w2.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "c16064f75ad6dbbfb8438ebfc68bb59dde23e9fed2c0ec1457ddfba3ad57165e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:14:05+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_natural_stories.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b6c44cd81b7255d91a6d08954964ade5a4aa116
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_natural_stories.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "b056520e18f9c1425014a6eaa4bc09c1b09f1b0d02f9f4d383d3c624f12f9983",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:22:23+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_natural_stories_maze.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a782acef47b2a63b8c48537b6e498da23e5a136
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_natural_stories_maze.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "5ebee08381a51ee920a625d40e1162050375cd3a5d07ef2772ae81bd488c81f9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:37:08+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_onestop.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff5d4e1ce2261ca9745683db123a0853aa5df29b
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_onestop.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "985cac240db5512f7daf7b78f5c53498e9e77d5c631bc8669d5e5aef8e32463b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:15:53+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_provo.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..f719e703e6d539eefe7274fa305e0420d249a166
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_provo.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "4d873c51d8f58d2076b97a6953c811ea495ab2bc765f9de12e568c502ac56d54",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:31:59+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_sbsat.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..b11cf531e7d0f473744caaba78d378a204971f41
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "d795b8e6731df2788d8b0acb4a525e04f917e22855979c3e480d894c89b4d2a1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:37:52+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_szewczyk_federmeier2022_n400.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..231f6263a7dcb12c2248f77ef540597301a63a38
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "2fb5bc62120bd28cd223054d918afc4e01d34845cd37571ff3b0a080b62fed19",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:59:55+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_szewczyk_n400.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ee7cb9e4af7bc7cbe5065bf28cf03397bee2398
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_szewczyk_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "a8eeaffc1c61cf7f94b7a1a4b57a27d02872554317e68719ca995ce2abde292c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T05:40:53+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_wlotko_federmeier2012_n400.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..54b4d3b4d043ebf9194045484e41270dd122c91b
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_wlotko_federmeier2012_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "2272f8b10cc9d640ba9e438887288a28517eb1bc50b53b1c1534595dd930076b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:49:44+00:00"
+  }
+}
diff --git a/manifests/gpt2_infoval_N100_M60_p0.9_s0_zuco1.json b/manifests/gpt2_infoval_N100_M60_p0.9_s0_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f35fff7670363fa9164a726de760575967f243d
--- /dev/null
+++ b/manifests/gpt2_infoval_N100_M60_p0.9_s0_zuco1.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "information_value",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 100,
+      "max_tokens": 60,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "spacy_sentence",
+    "unit_layer": "sentences"
+  },
+  "status": "completed",
+  "payload_sha256": "f2287ac2632dd52ed83014a1e9a7bba207bc4ce80c192e196c4872d87997db18",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T04:53:08+00:00"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_celer.json b/manifests/gpt2_logit_lens_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..ed1741142fa277c2b101deed8a8d4833541b7883
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_celer.json
@@ -0,0 +1,113 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "83cc34ac27636e73bac36d84b208faf742353ba6e8b7f1781a4d6966d70a03be",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:35:28+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_logit_lens_surprisal_tokens_celer.pkl",
+    "source_sha256": "25f59bdf817ecc158efc01d51aca2d1ded93cbe63644c64cb0bf600b2015df3b",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_dundee.json b/manifests/gpt2_logit_lens_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..badad88cd36f6d2cde140aeb239ae40d3f331fa8
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_dundee.json
@@ -0,0 +1,113 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d24b6bc7fcefb630080738a7b3157240b0b352f626093ae907a15538d4fe0d7c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:19+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_logit_lens_surprisal_tokens_dundee.pkl",
+    "source_sha256": "ff2452a0d19eb8c4f9a213faa88b5c6333cd11f177049e2ba0d6854e843cb87e",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_geco.json b/manifests/gpt2_logit_lens_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..f988c6a63bf6290d86ccfe5d4017ceef8d595ae7
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_geco.json
@@ -0,0 +1,113 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1e0ef42b88a69a94cd4f6ef0b8b7c2070e219f3930d937de0a31e27acdb9f65c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:45+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_logit_lens_surprisal_tokens_geco.pkl",
+    "source_sha256": "31a9b4669bbf260a8ed7a013aa84fb14097c866967562f074c306a65b12e1450",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_onestop.json b/manifests/gpt2_logit_lens_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..481decb5dff390aa8375f5f54e8a15b5ca4d4141
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_onestop.json
@@ -0,0 +1,113 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1ed6b64a42b5db79ee5594afe1a0fdad7ef9f99d48ddbf642b8908d4d5da1ea1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:15+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_logit_lens_surprisal_tokens_onestop.pkl",
+    "source_sha256": "21c36ab654cc4554df212f209d4e6b18bff3fe4e07aba1a17c28e79d1a9b2de1",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_provo.json b/manifests/gpt2_logit_lens_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..956b4e04fdee57b750135644f7d884b2f3301705
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_provo.json
@@ -0,0 +1,113 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8314e8185ed707eb919c4fce68e17666377ec4c9050c282f89b43477002d7266",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:14:46+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_logit_lens_surprisal_tokens_provo.pkl",
+    "source_sha256": "103fc7a90a9c630febb3e14efeb60b6f305b55a070d4047bc093ab5c8c0bdcea",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_sbsat.json b/manifests/gpt2_logit_lens_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f89b8709155c6ef954bf1442d470e8d7732c094
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_sbsat.json
@@ -0,0 +1,113 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e9164e045206fe3ae11bd9b98ef14ec3650a8029e4346a727691f16a9fb77a08",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:53:44+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_logit_lens_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "e75f23e80379bd200b6aaef75b9faae5ccedf91488174b1171d404aee0782f71",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_story_sbsat.json b/manifests/gpt2_logit_lens_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..4de61eb8f61f7e3fa5b1b93f52086e9214ce5b0f
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_story_sbsat.json
@@ -0,0 +1,113 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ba2eb8636cfdcdbadb25073e9026fb93e6cc7330ccb279a6b52c994e0a9dac7f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:14:46+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_logit_lens_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "be97c3751081a7aff712ef053724d155fe8b7ee64221a042380bacc908a6cb1e",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_tokens_celer.json b/manifests/gpt2_logit_lens_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..309f373a9c734f262b08312bb17f34e8956ec577
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_tokens_celer.json
@@ -0,0 +1,96 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "25f59bdf817ecc158efc01d51aca2d1ded93cbe63644c64cb0bf600b2015df3b",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:34:58+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_tokens_dundee.json b/manifests/gpt2_logit_lens_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c9722fd050d5e442c964363c82817a1297cb188
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_tokens_dundee.json
@@ -0,0 +1,96 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ff2452a0d19eb8c4f9a213faa88b5c6333cd11f177049e2ba0d6854e843cb87e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:08+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_tokens_geco.json b/manifests/gpt2_logit_lens_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c8d6f7f2222ae156fb4ccc09245e25eff21439c
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_tokens_geco.json
@@ -0,0 +1,96 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "31a9b4669bbf260a8ed7a013aa84fb14097c866967562f074c306a65b12e1450",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:25:33+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_tokens_onestop.json b/manifests/gpt2_logit_lens_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a80a7a0ee3de9bc6129a2f7226bddebe83ff242
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_tokens_onestop.json
@@ -0,0 +1,96 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "21c36ab654cc4554df212f209d4e6b18bff3fe4e07aba1a17c28e79d1a9b2de1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:04+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_tokens_provo.json b/manifests/gpt2_logit_lens_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..c8b7f4692b9fdc4e98c31beb66f23fc24cfbe32e
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_tokens_provo.json
@@ -0,0 +1,96 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "103fc7a90a9c630febb3e14efeb60b6f305b55a070d4047bc093ab5c8c0bdcea",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:14:40+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_tokens_sbsat.json b/manifests/gpt2_logit_lens_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..0282669f30f6db9a5ff5cdddf9ca43c1ff1fd331
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_tokens_sbsat.json
@@ -0,0 +1,96 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e75f23e80379bd200b6aaef75b9faae5ccedf91488174b1171d404aee0782f71",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:53:43+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_tokens_story_sbsat.json b/manifests/gpt2_logit_lens_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..52e63c6147517faea5fbfb16f1a676c4b6d1f25b
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,96 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "be97c3751081a7aff712ef053724d155fe8b7ee64221a042380bacc908a6cb1e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:14:40+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_tokens_zuco1.json b/manifests/gpt2_logit_lens_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f868193b586eefcf225bcf9e898a25d11e331ba
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_tokens_zuco1.json
@@ -0,0 +1,96 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "541eb642e1a83e1ae16819c1746931e33073207a5d1a7f225ccd5471fcf217f2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:17+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  }
+}
diff --git a/manifests/gpt2_logit_lens_surprisal_zuco1.json b/manifests/gpt2_logit_lens_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..2003d6cc008e26873310f01202bd712b82f83681
--- /dev/null
+++ b/manifests/gpt2_logit_lens_surprisal_zuco1.json
@@ -0,0 +1,113 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "logit_lens_surprisal",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "logit_lens_log_p_L0": "nats",
+      "logit_lens_log_p_L1": "nats",
+      "logit_lens_log_p_L2": "nats",
+      "logit_lens_log_p_L3": "nats",
+      "logit_lens_log_p_L4": "nats",
+      "logit_lens_log_p_L5": "nats",
+      "logit_lens_log_p_L6": "nats",
+      "logit_lens_log_p_L7": "nats",
+      "logit_lens_log_p_L8": "nats",
+      "logit_lens_log_p_L9": "nats",
+      "logit_lens_log_p_L10": "nats",
+      "logit_lens_log_p_L11": "nats",
+      "logit_lens_log_p_L12": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ea3fc557fb17aeb2a88b1e1464fbc3df59856c2a444afd8e4112d8eafe1b5644",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:24+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_logit_lens_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "541eb642e1a83e1ae16819c1746931e33073207a5d1a7f225ccd5471fcf217f2",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "logit_lens_log_p_L0": "sum",
+      "logit_lens_log_p_L1": "sum",
+      "logit_lens_log_p_L2": "sum",
+      "logit_lens_log_p_L3": "sum",
+      "logit_lens_log_p_L4": "sum",
+      "logit_lens_log_p_L5": "sum",
+      "logit_lens_log_p_L6": "sum",
+      "logit_lens_log_p_L7": "sum",
+      "logit_lens_log_p_L8": "sum",
+      "logit_lens_log_p_L9": "sum",
+      "logit_lens_log_p_L10": "sum",
+      "logit_lens_log_p_L11": "sum",
+      "logit_lens_log_p_L12": "sum"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_celer.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..53837c0202047fdd49ff1f8a62fb1b294ce7b5bb
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_celer.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7fed61054ca67072ff4fe12f6d7625ef0e0aa5ea8893ed05fa5c302e98056f16",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T03:05:54+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_dundee.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..a7d40aea864544c5aa3126dd17c0f7976c9bf4ac
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_dundee.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "91b20d8642635c3dc103864559eec1ca2ed0b219e510b0d4970a54add8a46d2a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T09:46:12+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..9405793fa00f75925a5ea7ce31485fce602451bf
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "aaae280de8d0097b817d5bb7c4ff6262c1482ceab3e3e1aed804955075d063de",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:03:23+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_geco.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..eca4a56dd25d768af1d5b2c0316e9e8e245cadf9
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_geco.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "1b491ea204807c70051a644103abafab2124ca7e4a47dbcd5bc84b08e04425ba",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:53:01+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..73bee6119100685275464f526b62d0872da1bec6
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c6588543fddd11a141e18676f16dd18390c0b0ddddfcba83d5453cfedf79650b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:34:13+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_natural_stories_maze.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c01e3fa2c9e539ca636c7491388ba02d3bc1313
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_natural_stories_maze.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "9c86922fc133ff045746da8b6526a325a2ce6683d34b06c25357fd4e9ded546d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:18:01+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_onestop.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..dd10f140a19652553dea84b4ed152a2431bf6bd2
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_onestop.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a4366ecfb7842205e08c389127dccc998a563e25c41f80747b6af03fbf469b93",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:24:02+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_provo.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..91e2aeec333f979a4b563aa48d033088a558ac1c
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_provo.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "24e80010ec7c8698b74be97b212a837150225907688918834d54ee693c615c27",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:54:19+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_sbsat.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..792e47694f129fd574ed9ec7124e08bc433486c2
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f30c3f5f9212e75003a749c8dfecb055ef3dfabd26f186f63ed0754ef0adb2c7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:28:35+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..94814c0519dda69614be1b2a95eb0003e399c242
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ac6631b82277b8816306fc459b6f9039842d8bab375ecf2199ad3f23a6f28133",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:26:56+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_federmeier2022_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2d0d9f94d865927dcff1c2ae06f3cf926f71db2
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ed751e1e718a1fb66296256ac73f7c9acb39b7c57917dc69c607edcddca2a347",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T08:07:11+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..186645f323ce0ebd9bbec3cb4d552767febec8b4
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "3f8a7dc6a3e12508dcff87d78444362e0de18edf568b20c1248ac70e796616a2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T08:43:52+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..f80f05c0c768fcf447e51da99c31aec85e9a72fb
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c544e932663d94d6d300d247d8c2331fafc811e8298267bb1b049c4362161cfb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T08:27:50+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_celer.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a68f2002320bc23f7d7bc89f0adc0d836320160
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_celer.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "832db1cf0a81aedb1a549cf99a66bbfb3394e038bdf59ea424ee6a199e38656f",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T08:50:07+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_dundee.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..caf781b906b14a15463c139b952efe8faff78184
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_dundee.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "07c826593f751be91425c07d4b88916ae1e93a40d88c26856e32a64acdfa5c37",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T10:15:00+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_federmeier2007_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..72cbd105aed6f1413ff79e862e2c0f4430f1713d
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_federmeier2007_n400.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ef5770ac4d697672ce534b4637b7bced58c679aeed0dedc25819e9454eb0d407",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:13:20+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_geco.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..6dc9f8988c95e784b3f99627e4be87b28e3598f2
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_geco.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "032c1ca889480b4e2e7ba2f92923552ad43a8f58d80b4838ccc25f26c5f19d06",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:51:31+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_meco_l2_w2.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..cb393e2069a38ecbc09e9feec2ed03814f23b41c
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_meco_l2_w2.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "bb18079871ce4a59a7e8e07a6664d44ebfe0389800edb76a53366add293705b8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:36:02+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_natural_stories_maze.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4615cabb0c3921a2a123e49302f99aec51c677a
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_natural_stories_maze.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d7f7bc41aeaa0be409b99e2ac10b41852f0d44ccc2a9b41013a8a4c84a6d692d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:31:01+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_onestop.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..99cc0287b88d65221803549610259d5c167cc3aa
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_onestop.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b883bec72b96a0da91c7e8af19152c41f333fc59cd900bae3ea61d03f79dc17a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:06:48+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_provo.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..50f83afe683979428b7eb4d916d429138b94e08f
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_provo.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2198d700a88113f3a3f3b61d80df2924df1d1720a9ce7f898f5757f0f28b5e98",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:11:05+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_sbsat.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf4b36af173940052fd177e3fead5539f2d05f92
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_sbsat.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0d0d7d25c78bbcf225f942696306e50b9fbc7f53ace679256dd584069e7f671d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:30:21+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_story_sbsat.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..e9f1509a40fcd229bc636d0d45679af3bd4379be
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_story_sbsat.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f9b884f29489cd807f4b5e6275d3c65edf179f6273b800654fa5b8f759d69bbf",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:26:16+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_federmeier2022_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c400055071750ddf04c6574bb9cb8e4e978601e
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f908bedb4ac833b9767dbeba038630d6007c2fbe9592482844468e30234767ab",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T08:12:38+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..202d670f3ff0fa95d1a8727078b56dadf0931b7d
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_n400.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ce06811cdc924002e4a8b4378afdfe7dba382ea4df6c2075fd039296341accf6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:00:45+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_wlotko_federmeier2012_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..1fc371fb13e59c5698b6a1affd8c71d5c2927012
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_wlotko_federmeier2012_n400.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e9a1f40e064a64fad5cf88a852428b1aedb200f1c80a22b0eb041a0c05a411ec",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T08:35:28+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_zuco1.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b5b09af206eab02165ba3665b8896201fd1b7fc
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_zuco1.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "02ed22629ec0bd697bd8ccb2f11a017b4f15ba4aaaf1178aaafe1a0e17edd767",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:04:00+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_zuco1.json b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e262ebc6fc68ae043354738a61b8fb793c7d5e1
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a0.5_s0_zuco1.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "1337469b80ffe8cfcaf5271c253fa143529cd6adcb104904b98c9afd86138ae4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:59:11+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_celer.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff546d76b0807d7de9e2fd74e0ccfae625a87da0
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_celer.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "32de787e059eb44bc9af7fdd037cd711fcbf8ee9462867ae32de49cfc846a5c1",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T03:43:02+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_dundee.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5271ef19a38762a0c15b34c6808dd2f569c21b5
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_dundee.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "54bfdf3ed7b307da36c5d2c8dbc1cea2dfef9a0dc577b2bed378efc18ea8e07f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T09:29:03+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb8cbce8ad5fc3dd6596d1e8cfdc35733c7f8af3
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2c4564f116849058f30a7ba8afd7354d24064985bb474d35d82f19917c086b73",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:12:24+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_geco.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8a15a540c2c5c69d4189e14061276601ff05ec6
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_geco.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d9b43b383a9bddbec4d4f946db9741915b6c3f532a8ab2ce5f76b9554dad9cd1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:52:23+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..e125f94218f84d73f24e68beb5b22392301b7303
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a01b00f0cd180d853180eb98d78e474cb32392c3aacd1f3e5e74dd7565e31925",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:37:49+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_natural_stories_maze.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..a70c519703544ed34fd98639f2d1077d4c74dcce
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_natural_stories_maze.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8199ff51d2900c30aae6deaea62634115753889b9b8ffe1ceb85bb471c2f2801",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:20:07+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_onestop.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..89c6b5d94aa1819ca766394bca9777d362e1ce5f
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_onestop.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7bfbc5669196a1f06425ad3c47e4c2496a7c3bea39e976a519c09ec301767577",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:24:42+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_provo.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..5566eb91502f01df6120dcf6d3538b631a8c3f90
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_provo.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "109f9643b119a960d3d7909dd10f086fdbc0e59801c1d0577bdbb6ce6dcfe54f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:14:59+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_sbsat.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5c37ca08f4dca7ed1bcb05bc0dd2c651a7ae1e4
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8da423571b1ed8f4253d3dbaba49d0020ec4752629364d2cb9968b0c0bcb0b76",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:28:56+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_story_sbsat.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..80c1cde7ae892c798d9c4f9e9cf35399a96fa311
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_story_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "bcd48be9ef9f1ef10b20a9979cd12781c2c7f035ba099cad17e7ce536024a605",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:23:37+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_federmeier2022_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..665171dc7dcf7db6fefd2287d343712fb34d063f
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2247c512d91a10a8187c93bad29ec06f08255fc5ed00424ef55679700d784e94",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T08:08:53+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..2645128ecc4bc174f6f92a9cc289b90047b19506
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "95ed576e477c19f4854013d47f6a2b14d527dcd88d9230643a0a2c09d2e9d610",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T08:48:37+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..c790ba11183a0f5053192a1f61397d4993bda570
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b6ca8f73c3b6fe7519a410a926f2db92a419250b8cbf879fb504a282f319876d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T08:35:07+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_celer.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..4382665ee42f23a82e111c39ca750c890b7033ea
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_celer.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "aa400723e8bd9429152cee24ee2982695cca1a8235853770c41731f5cc133776",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-22T09:27:40+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_dundee.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..6fe059b4f2813fadb2de06e6ed29717f1ac115b7
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_dundee.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "665a39668e167a7d402f91121c92ca9a13560c1a271df8ae3722f2ba3e617911",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-21T09:53:28+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_federmeier2007_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..52f2d0ea8aabbcb3a814a4d36bcbfb410df44459
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_federmeier2007_n400.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0aeb60bebdbe3bf3a5f051ee8fcc39c3bdbddc02e066abe93ffbc22fb8449ade",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:19:56+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_geco.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a2cb375d67615f69a1766335bfb0ca12b33470a
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_geco.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2422a940fcd49a9722672e01535f783250c59f816bc3bb1cd6a81781180167ea",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:52:03+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_meco_l2_w2.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..26549c87fb6a66034fa94bb57a684e8a6190ec25
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_meco_l2_w2.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c6fa32a25660946d046e0c8502bd6b634760327158d0a0da5dc992c43667b6a8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T13:40:03+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_natural_stories_maze.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..13043283652def403e03fe8787e7c883998e7988
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_natural_stories_maze.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5d6e2369360d80a94d4c6aa225014fe736533517b48446d1e689967d43ba3e29",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:33:06+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_onestop.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..76f1c34bf5ecddc76bd3ce04653614873e5691d6
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_onestop.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "fb83040d42dca266a6d30f9e712a1c2a443c2c7f152b37fea3505ff73d512791",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T18:02:25+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_provo.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e0e5ba269bafd66ef60786f657e0f1ca67781da
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_provo.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "818b209988a1287161ecc8f6713d210ba0a6933456a8019ea5d6dde3fb76d4a2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:17:37+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_sbsat.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac7862c7ba587ef8a75ac20d8febf2903fe8db95
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_sbsat.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "fd22940e56865ec14ce46433e6e8e24dcc382dfb7a92a126a1ffe7c42a48dc24",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:32:59+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_story_sbsat.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a79b42076d5b68ddb5810e6e6bc3fe8a29984ca
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_story_sbsat.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4f03904c2b41819154a0f3a0ddbb0b2625d3206af06f20b6c7811886960bfec5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:26:36+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_federmeier2022_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..96bb7c62be30b670d76bf9bf9429180cbbbc149e
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7aaaf0089c06a8af65be5334750d968929ac1bce1ddab138947b97b0e903a189",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T08:12:36+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..c18120135983108977f3de65d1486db8f298ca69
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_n400.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e432cf382ef541df4fa174aae061f78b762da7adf342f984a4130f6997ea80c9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:25:10+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_wlotko_federmeier2012_n400.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1bd73bf6244fd0c75fd3bd269b4f38cbbe04ea8
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_wlotko_federmeier2012_n400.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6f558a211d48c5e192b1489c915ba5f1a3d7ab2364a109b95fce5843665b802e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T08:40:06+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_zuco1.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f9f2ad4c228e4552207e14cfaaf5b4b963e03d8
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_zuco1.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e7e00e6d160bce54bfcc7230093858d83f9af54e9da4efe6c93b6c4510290e64",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:05:49+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_zuco1.json b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..2664907acdbfbb090af1d4f0f19f2a17dea79b37
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N512_M20_a1_s0_zuco1.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "1460f54face3048a50d279c3068ab833e2984c4ce760a35228f63b98120052f1",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:58:27+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_celer.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..35338ffb01812277a7b7f365da5838b605c80a68
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_celer.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "aa0d3ae7642ea888fbf2bf293b16151e6feaf814c61f140c649edf7e7c9132a7",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T20:25:40+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_dundee.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..a75a4b923d44aa77167f17dcc32a5e4e745928f6
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_dundee.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "894d341b16348cd7cd09cb69bb02d9380d1eeffb2a7ae7c4ac9c3c9ed1b5c626",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:00:47+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_geco.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..1dceb940f5d11b07a5e7ca34d9aee029a55bb6df
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_geco.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c366d0a803946e7f50c7a73c520e4255587ea8fe2c026f84be547c9de8c6b4f6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:25:40+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_provo.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..d4b40b3934b49c3b8874e13835d7dc7127bf730c
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_provo.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2d640cca7f982dc0d35dd47e4944fbfaa22306b159bf41d7cd729ad8e9f907cd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T07:36:21+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_sbsat.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..689a08f08fa10ef6639fbce27e2b877fb250e8b7
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4fe593717298b8d4dcb31dc8854a07562e95faead3938eeb92323d5473b3e487",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100-SXM4-80GB",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:22:17+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1cc6a49affc784d03547fd5803416acd029cd1c
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "bc067e0d77bd36d6edb759bb29ea790222561f92181e96b535f1e1d5255a04ce",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100 80GB PCIe",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:34:20+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_celer.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..0da2cc60c8ee924bdffdbeb19c6faf1f391e1329
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_celer.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "645c282153f1764d9706ab156af66c328efa7ae0523f10b513810d91435646f2",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T21:08:43+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_dundee.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..195cc77c48f9ad6891ec577ae834f4908535ded9
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_dundee.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2eec9e428f5cd62c81c4b81f13f6e2f4cea46fe43b1a80a90173dcaccfca3198",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:00:40+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_geco.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..34e173bbdb03503b6fc862ec3f98064b25f1026a
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_geco.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0c6a34f334e68ad5ee8228c976111236bf6b23ef576eb85de7eac640c51a25e9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:32:54+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_provo.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1c0e4d550c866b072ac3d939fa6e818d3d2bba0
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_provo.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d1fee659ce316d7d4dbfa917fc8411bf92407f20a2ed2e8b9df771f464ee4f18",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T07:38:08+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_sbsat.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..e7629e18391f12e2bebdef47fc02e0771dffef88
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_sbsat.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "3ff26f490529df4847d2505ebf1900191b1102b8b355944b86ba04a4276d3d70",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:25:37+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_story_sbsat.json b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..34aac21bbe6babaff6c201af790acc91ff3cb824
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_story_sbsat.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 0.5,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "815fb579fd7973a755814b9c765b5835eade5538d363705508cd54598565149d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:34:51+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_celer.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..cc6a054726a1243ce692ec470ba19545969c537d
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_celer.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4f0182f8768ee18a50c52294525313e0db2d77f15ff915db6b4649fe468fcaca",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T21:11:22+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_dundee.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..706f16e996e61c37da4f12621f2464a6c5dd176d
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_dundee.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7f972bc8cf84d6de881c6239bb3f7a35676a25bed82be54b70d765a994277675",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:01:05+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_geco.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..3baaadfb2a35ba165f1eaa5276b933714d8fd74a
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_geco.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c79213645ccaa8818288f3a70c00ab75ac35372295ce6997e8035ff6e0616045",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:25:57+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_provo.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..514c5ca42e69a8f26ad415e7593b4d5c1a903a94
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_provo.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "603f4033cc12b3b385e0ecc396784b828afe2355e1b09de6856e51ba2b3cdcf2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T07:39:03+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_sbsat.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..730eedc45b31356e1b9db8bfe4eceed622df22a1
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7af72f62b13c04ce483b18c5fabd60425d5e5c2e3edca7d930d6378c1a5a58fc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100 80GB PCIe",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:25:31+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_story_sbsat.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..1872bd4cf5a0ba7111bc59c0af6f083b336fe2e0
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_story_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c28c2575639091dad3045ae7abc647a2c821eebec61b2e938c8124140a89f981",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:34:23+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_celer.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..ccd023451bf563474afac4bdc1ab9fa1d99bbe69
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_celer.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7682026edd711aae56a17d414bdb498a6bdb7087be7305f150a4fb1fabeb9559",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T22:03:39+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_dundee.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f43fd37b64832397c413a46c41bdfa712213397
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_dundee.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "84751d0c4394b04c3e794eed49ab17761afccc1511edf5f516ebb07d6e284ca6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:01:07+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_geco.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..f9efd3e2bcb7ae2d4b3dd29056f3d1fd911a98fc
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_geco.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "bf5f6c48063c5de0fa1071f9f1cfaf0701bd3713194b0260f235d6f04ca51280",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:31:38+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_provo.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..cad796aa1ae12d0f42372dd35cb89e1956fb4d2c
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_provo.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5df2c636234e7d498490d5ff1d7fe512c3db179ec60d4ba900689ea1104ad970",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T07:40:19+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_sbsat.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..37752bf3ca6eaa83ff4601c91e0ad5f3171d999b
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_sbsat.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "568a7a4302c24f0ae0801b561e80b1365535414523da2fca295e11e723f0596f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100 80GB PCIe",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:26:03+00:00"
+  }
+}
diff --git a/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_story_sbsat.json b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..45c7b86c1866af28a29e5833aa954b550df9b06a
--- /dev/null
+++ b/manifests/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_story_sbsat.json
@@ -0,0 +1,79 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mc_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 20,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": 64,
+      "word_starters": [
+        "\u0120"
+      ],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "fbc35b0ca489a96c0bd1f13bdc1e5227add95d3aae46687373a261d254cd9d5b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:35:41+00:00"
+  }
+}
diff --git a/manifests/gpt2_mmc_unigram_surprisal_federmeier2007_n400.json b/manifests/gpt2_mmc_unigram_surprisal_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..20ec1fad85ccda196f96aaff9e43e28d0adb9e8e
--- /dev/null
+++ b/manifests/gpt2_mmc_unigram_surprisal_federmeier2007_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b06ddf19d8ac68cca1ef083b6537b905a11ce10f14ee47f16ceebdb0c308912b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:55:05+00:00"
+  }
+}
diff --git a/manifests/gpt2_mmc_unigram_surprisal_natural_stories.json b/manifests/gpt2_mmc_unigram_surprisal_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c4f50fe1770780ce8ca0a9a92e1a0fe18231d42
--- /dev/null
+++ b/manifests/gpt2_mmc_unigram_surprisal_natural_stories.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ace1c60156093180c0c1043006ff81c9a450be7f720c53621a195758b057366d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T19:57:18+00:00"
+  }
+}
diff --git a/manifests/gpt2_mmc_unigram_surprisal_provo.json b/manifests/gpt2_mmc_unigram_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f431c6bae40eeae45039a7945255600e9d4361c
--- /dev/null
+++ b/manifests/gpt2_mmc_unigram_surprisal_provo.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ddb48e5b93dc50b4c5fc5e17e1a59e643edfe496ac6f17d75471df5a5baea72a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T12:31:09+00:00"
+  }
+}
diff --git a/manifests/gpt2_mmc_unigram_surprisal_sbsat.json b/manifests/gpt2_mmc_unigram_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..069d68ba61f11be65f93fea1861cf815ccc367c1
--- /dev/null
+++ b/manifests/gpt2_mmc_unigram_surprisal_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2e01f9d542e1ee0d8d9783b0de5f547415b0baeff4d6248b0fd6da5398b2c642",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA L40S",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:04:30+00:00"
+  }
+}
diff --git a/manifests/gpt2_mmc_unigram_surprisal_szewczyk_federmeier2022_n400.json b/manifests/gpt2_mmc_unigram_surprisal_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b0313bd733ed34c013efcc42a83cbd24bd2fb01
--- /dev/null
+++ b/manifests/gpt2_mmc_unigram_surprisal_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "22f6fabffee7bff50a0ad22acd9e76c71e3ea102bb60c2f6c1e2ecf303a5fd7b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:29:56+00:00"
+  }
+}
diff --git a/manifests/gpt2_mmc_unigram_surprisal_szewczyk_n400.json b/manifests/gpt2_mmc_unigram_surprisal_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..c418ddcd334a6c0b128b4631c9deb0d9e7fe45d0
--- /dev/null
+++ b/manifests/gpt2_mmc_unigram_surprisal_szewczyk_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4fe8713c01662a60665da66d68629ed8f1c5cb194184612138e4d03b88f85fd7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:03:24+00:00"
+  }
+}
diff --git a/manifests/gpt2_mmc_unigram_surprisal_wlotko_federmeier2012_n400.json b/manifests/gpt2_mmc_unigram_surprisal_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..1513771761420e5e71d117d1e38fee53fcea4de4
--- /dev/null
+++ b/manifests/gpt2_mmc_unigram_surprisal_wlotko_federmeier2012_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "mmc_unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "22ba584916c0d73cf213744de3679a3b09e08eb55f3be492566870f081e5aa3c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:14:33+00:00"
+  }
+}
diff --git a/manifests/gpt2_modelblocks_surprisal_celer.json b/manifests/gpt2_modelblocks_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff06fea23fba74f9170f2e2c9cc9a749dc8100c5
--- /dev/null
+++ b/manifests/gpt2_modelblocks_surprisal_celer.json
@@ -0,0 +1,78 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "modelblocks_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "clark2025wordentropy",
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e3e8a68aa16edbe349019b82c26b89e615a3422df7bc99f780270f23eacddecc",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:50:04+00:00"
+  }
+}
diff --git a/manifests/gpt2_modelblocks_surprisal_federmeier2007_n400.json b/manifests/gpt2_modelblocks_surprisal_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..e91cce0c5a09066ebde98a6a460610c496c6f2e0
--- /dev/null
+++ b/manifests/gpt2_modelblocks_surprisal_federmeier2007_n400.json
@@ -0,0 +1,78 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "modelblocks_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "clark2025wordentropy",
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "cfd5ff60edca20542290f6605d0b1ec8d5e2390aa32394285211a5f63de16a12",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:21+00:00"
+  }
+}
diff --git a/manifests/gpt2_modelblocks_surprisal_natural_stories.json b/manifests/gpt2_modelblocks_surprisal_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..f28b322978f2b9795b951eb9bb1e03f7a8d10146
--- /dev/null
+++ b/manifests/gpt2_modelblocks_surprisal_natural_stories.json
@@ -0,0 +1,78 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "modelblocks_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "clark2025wordentropy",
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "935e297e8dca9d96bbd16029315127aafbf809ea647ba6af9da8aa91f11842ff",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:10+00:00"
+  }
+}
diff --git a/manifests/gpt2_modelblocks_surprisal_natural_stories_maze.json b/manifests/gpt2_modelblocks_surprisal_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..dde0c7512fe2304ce4a6cfaedd363e6448d0dff0
--- /dev/null
+++ b/manifests/gpt2_modelblocks_surprisal_natural_stories_maze.json
@@ -0,0 +1,78 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "modelblocks_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "clark2025wordentropy",
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ccbd658ca6ad59bbf30ea64b459d1874090968914e43f819658741e70b599a35",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:20+00:00"
+  }
+}
diff --git a/manifests/gpt2_modelblocks_surprisal_onestop.json b/manifests/gpt2_modelblocks_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..d0e0f7f1458ad12dbf9f9272afed0d4030790d51
--- /dev/null
+++ b/manifests/gpt2_modelblocks_surprisal_onestop.json
@@ -0,0 +1,78 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "modelblocks_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "clark2025wordentropy",
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d2d72d56e93430962ac71119a61f2009d6baf3d1511dc635087c90fb210d2b00",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:03+00:00"
+  }
+}
diff --git a/manifests/gpt2_modelblocks_surprisal_provo.json b/manifests/gpt2_modelblocks_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae2a8fbf50babeeab473058a9ba846e79d9d5633
--- /dev/null
+++ b/manifests/gpt2_modelblocks_surprisal_provo.json
@@ -0,0 +1,78 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "modelblocks_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "clark2025wordentropy",
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e86ed6374edb5d6a8de860e31a660bf128c49073079cb39b68991831e215d0ed",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:10+00:00"
+  }
+}
diff --git a/manifests/gpt2_modelblocks_surprisal_szewczyk_federmeier2022_n400.json b/manifests/gpt2_modelblocks_surprisal_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..cbe36a3d89298059be44667ac6f31c14008021be
--- /dev/null
+++ b/manifests/gpt2_modelblocks_surprisal_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,78 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "modelblocks_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "clark2025wordentropy",
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a8fb6161b33f5fb808161e270b5ce754a86e6b1b48b819c27b7081ea0a51e6bb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:22+00:00"
+  }
+}
diff --git a/manifests/gpt2_modelblocks_surprisal_szewczyk_n400.json b/manifests/gpt2_modelblocks_surprisal_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a7f44988c1c7562e70c42e291b87b8b0c6dba64
--- /dev/null
+++ b/manifests/gpt2_modelblocks_surprisal_szewczyk_n400.json
@@ -0,0 +1,78 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "modelblocks_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "clark2025wordentropy",
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d30bdab4910236b78bdcba980f539828a92b438a7bfff499e66e69aba5116695",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:43+00:00"
+  }
+}
diff --git a/manifests/gpt2_modelblocks_surprisal_wlotko_federmeier2012_n400.json b/manifests/gpt2_modelblocks_surprisal_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c3b90b5c7ea4dd5ab0b4de125088737431fe2d9
--- /dev/null
+++ b/manifests/gpt2_modelblocks_surprisal_wlotko_federmeier2012_n400.json
@@ -0,0 +1,78 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "modelblocks_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "clark2025wordentropy",
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2c88bc68c2cddee80c4d492b7cb40fc2a834395553af915377537a21dfcefe09",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:15+00:00"
+  }
+}
diff --git a/manifests/gpt2_modelblocks_surprisal_zuco1.json b/manifests/gpt2_modelblocks_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a2da798a8a9d53fbd8e6787109730a20d474c6b
--- /dev/null
+++ b/manifests/gpt2_modelblocks_surprisal_zuco1.json
@@ -0,0 +1,78 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "modelblocks_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "clark2025wordentropy",
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c5e8827d19d25e90a8ca721df510345d37d4f93660f923908747d2985a5c1fe6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:46+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_celer.json b/manifests/gpt2_next_word_entropy_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d33f5903381b345c7e81eb515fac6ccdfb09b55
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_celer.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "acc53e7fa747e22f2402eb705c78a6cdc4a5808af55d9872faf21104f0e2b47c",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T15:54:50+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_dundee.json b/manifests/gpt2_next_word_entropy_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..dbe73f716cd6d3a2fb453caa9067a2baf851c3f1
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_dundee.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "73aabacbaf6dff953b1734ba8ece6c753b4aba3cc9e85f835f71b728d95f1e6b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:27:44+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_federmeier2007_n400.json b/manifests/gpt2_next_word_entropy_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..a974edb01e5189b433356461b2d0ab560fd0fee0
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_federmeier2007_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5bce3b9504a6aac2962c12020d19a7b7b61c4db53c93764fa342ea1a1eb06791",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:10+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_geco.json b/manifests/gpt2_next_word_entropy_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0286e311a97f6fa34497def876264fa2ac30ee8
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_geco.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "40161ec52e2aa2333a726b4c82c2c89f490668e8b289eeed70dcda2892597303",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:45:50+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_meco_de.json b/manifests/gpt2_next_word_entropy_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c3d5b594f993de2304b08c31a6e4091191baefd
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_meco_de.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "63381acd186301656b5de883ee2a88bbd9709c1d34311d793917e5aeda64a17c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:44:20+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_meco_ee.json b/manifests/gpt2_next_word_entropy_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98517a5142edc612832b7da412f108a02196d76
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_meco_ee.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f2694b98e2f5dba1681b9b0748c3030979c69c274c048d682c101a44f3771cea",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:07+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_meco_fi.json b/manifests/gpt2_next_word_entropy_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..8fb6b4ac9cff0a9e2476b182c2128fd5a9ccee65
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_meco_fi.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2498b477992dba202ff72bc4beb4d1ec97b2f9452b629e4034422caba390cd5c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:28+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_meco_gr.json b/manifests/gpt2_next_word_entropy_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..33ede851c8b93e93fab8f2598a03f428b2576e9e
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_meco_gr.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "48c4663b799d6d1ce27b1a756b809622f2949329e94aeea302066ee0464a4439",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:28+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_meco_he.json b/manifests/gpt2_next_word_entropy_meco_he.json
new file mode 100644
index 0000000000000000000000000000000000000000..de17836eb2e26d7ae8221ee7c5249e540166118b
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_meco_he.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "he",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ed49348beb7e184f5d9baea8291cbdc0e8f6ece7dd575d993488213cd37826e5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c67cccb428c27d940e630582e9b8468f4355e9541ba0a42983d24bcd0665173c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:28+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_meco_it.json b/manifests/gpt2_next_word_entropy_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b8b23347cfc7d8c0a72dafbaf0bbbd8ffabbca1
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_meco_it.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e0be54fc1594f839e4f93d41b52b8ba4d187e8a36ed05487d8d9b2f72ffeb3fb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:15:38+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_meco_ko.json b/manifests/gpt2_next_word_entropy_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..73ba823be5779e2b5d63a35e731a8ca48a3bf290
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_meco_ko.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "3cd60e76ccc680eaea63dbdd75120d1a63abc23dd10b68931e20f4b64eee786f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:45:42+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_meco_l2_w2.json b/manifests/gpt2_next_word_entropy_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f7bd722452f5e015a046095ef558426bbcb9c8b
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_meco_l2_w2.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "844b4f10b977dc8e8eea067da965d839e741fd127d5a6154525c1e1734435745",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:53+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_meco_ru.json b/manifests/gpt2_next_word_entropy_meco_ru.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc11ecf36814bf8711e8fb9caf0ae559654831e4
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_meco_ru.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ru",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "56ee0a7928fd1a4a985a1c9e9377c928263e569df23ce0f3e0961329e7e6462d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "1098f2dd2619e5c69d425bc31eac5381a21b00c58320746d72b2410422151cea",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:49+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_meco_tr.json b/manifests/gpt2_next_word_entropy_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..c33bb667245363876129190635cbea64e7db32fb
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_meco_tr.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b70709b775d421c33777793577049e6abd11df340f8cfe232d5a8795891d6fe6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:49+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_natural_stories.json b/manifests/gpt2_next_word_entropy_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..585de1e335b292194cf3c279c085999ce11492c7
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_natural_stories.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "fd3ae39b0a032c2747c6cd7ce97bd3e9bcf764783e5a084a4cd9b0c18aef1f99",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:38+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_natural_stories_maze.json b/manifests/gpt2_next_word_entropy_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..3fd904fef54d6ffe5dc5e48687a6f8dcf00a1674
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_natural_stories_maze.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "676966434fe3575939dcd046c8246c9222cc93ea672d0bad667dbfa0271604a7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:46+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_onestop.json b/manifests/gpt2_next_word_entropy_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..15a6a114da4518920360b461c5ca818b83496fc1
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_onestop.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "bbeec46d61056eda697c8d3976c84970974a423d95e60b0e4cf8fe8b3de8e755",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:28+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_potec.json b/manifests/gpt2_next_word_entropy_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..703c7f7ca18a90bbec80c2eed967fe3e1d9da7f1
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_potec.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e6f840e768ae4885c353bda1725b572ab5c856cb68c02cedccdebc5159331fab",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:02+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_provo.json b/manifests/gpt2_next_word_entropy_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0485a438d0abec0f5eb0c95f2364e2b3d48ba28
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_provo.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "279b57c9d5f152f8587003f9030bb61ba4bc8a3ca7e88c8355195c18c4f20614",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:55+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_sbsat.json b/manifests/gpt2_next_word_entropy_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..31391f66545819770655455c9ddafbf22cc16870
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "860606d48c695df9e16b749657fc3a7fb507a7b0f5e13851bfe1c6dbdd694574",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 8000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:29:25+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_story_sbsat.json b/manifests/gpt2_next_word_entropy_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a2d6acbc989290142309b5600dbb6283fbd0088
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_story_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "cd699b1b731494d31e57a7022565de4c0d54b0fabe932aa31c7c3a6f3748fcd9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:46:50+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_szewczyk_federmeier2022_n400.json b/manifests/gpt2_next_word_entropy_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..95f92fc9ffe949a50cc739f5f7c9d72b52000bdf
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "127f6da1059dea0399c75924385b37e7ffce4d46196b5cfefeb48386317b091b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:16:46+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_szewczyk_n400.json b/manifests/gpt2_next_word_entropy_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3deede8c4d6691ef7d1673f0d3d448c3917abb6
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_szewczyk_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "99006da224f8e0f51b15c2a3a0184593388a46de67424725950add8e1574d438",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:20+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_wlotko_federmeier2012_n400.json b/manifests/gpt2_next_word_entropy_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..802b78ced4271c968d805acc68f6c1461ae06bcf
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_wlotko_federmeier2012_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "60743defb98aaa1dc31e3fb4b0ebbff3143b704e70e8d57fc9490cda5bfb60c6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:17:09+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_entropy_zuco1.json b/manifests/gpt2_next_word_entropy_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..873df2238d2f032a13be32ffceb629de75b1d4d9
--- /dev/null
+++ b/manifests/gpt2_next_word_entropy_zuco1.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0b13c3b3f6c645f56e2bdfd06e4881b5fc1c9d7f3a263e04dcb24111aea30582",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:28+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_celer.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f1bdc760b694543512117197263ff8a852b2093
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_celer.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e7ded646352fd605e8679fefd165cc2169f3b9f51db8957dc8ccf37c20528ebf",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:03:40+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_copco.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..d659e7a245d422bc790380d6a31e72be3d100017
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_copco.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "7f21c1d1e28331c8d44b4d098dfcf6896542e6c2ec9140751f7edf2e6df08738",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:08:52+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_dundee.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..52cfdcb4921f8d9db4b38a625b4e8bc9f966cdd8
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_dundee.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d708dd78e533dbb78d0869ce09d64a730047af19ea12010a266f6f389a94ebe8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:27:37+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_federmeier2007_n400.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8a9b02f496a70a71002e821a0c8ab39c92cfe20
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_federmeier2007_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f2dd33cc5bdf061a4f26c68540ffe6fae3f846a54309d6bd205d7bb05a52e3a5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:00+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_geco.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..5be3f3d858f5db6468efedb4d7b302a906573ca9
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_geco.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5e19e8af1306a5a1fa589fec6b924c80420914b3a69cbe5bd6b635fb719464ee",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:45:52+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_de.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..dd980c912769641f34d2401469b559a171fdb938
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_de.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b0af9f345c813889089e2929157edea82c7948139f277407ba7bedce39d3922c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:44:39+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_ee.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..286a660b925d69b09a60bda6379af71dd27f63c2
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_ee.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "b29047cf6f5d693c0222ec9d289c0433d69aaadd68b6f556a65440edf2800e06",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:07+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_fi.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..c018fb69b87b214a885ab8bb7062637238dae276
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_fi.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a380243bc1ef6fd785124f5fa1f5620636d4a50ffa879665275c7db1a7052285",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:07+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_gr.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..8be2028f2ad6eff9b73fcb12c3ebc4c51f9fe676
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_gr.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "178591f107d2bb564e42374f28c3485bc26f7f18d4fc506c86b48e9f38c9b38b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:02+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_he.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_he.json
new file mode 100644
index 0000000000000000000000000000000000000000..5c204bab91e8bf197410accda9ea22bd68c56a7f
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_he.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "he",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ed49348beb7e184f5d9baea8291cbdc0e8f6ece7dd575d993488213cd37826e5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "11fd5b0b15dc0a0dea910618e16e7d7d611f3b04c7c2d11eb29128b640b62d00",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:34+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_it.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..acf381a5d0d6af8f600dc635ec43e0cd9f9c8b5f
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_it.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "bea208d61a818909d16f0a12819f9fd69a83cc836324f050d31644ef42a19c9e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:34+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_ko.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b4feddd3d3d95b5f38d7b0ac1f4188fb15450bf
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_ko.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "1212d37ddf30af3825f6a0a30cdb1a2f9c2d1cef972a1356b6a92932430b250d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:45:43+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_l2_w2.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..a08581f948bd47fbc24d52735ca9c9e211b58cbf
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_l2_w2.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "caa9c34b8e699666e638f95f67a4d464a73d1cdf584efadfe1a4c1c27261a00f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:26+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_ru.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_ru.json
new file mode 100644
index 0000000000000000000000000000000000000000..987322ee4b70db2527872fdefac1c3e11f13c971
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_ru.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ru",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "56ee0a7928fd1a4a985a1c9e9377c928263e569df23ce0f3e0961329e7e6462d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c87023f02e6cdf7a13af3fb662eeb33625a9ce949fd7a57c050d563418084d8a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:29+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_tr.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..173a3068fddd33642f899c345afaa86cc3cba477
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_meco_tr.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "305b1058aa1a6b2d0e7e0815df2f37afa1ceef73fe20e0561f5304981a3b34d5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:35+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_natural_stories.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1be70c85235de059524810ea0035a788039df52
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_natural_stories.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e61894bf966d13894484a06224232ce473cce1f163a25aa074f032b0a54c04e9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:38+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_natural_stories_maze.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..4708e8140efd4af17e5485125e8589f25059c246
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_natural_stories_maze.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "cad916bb47a65eef0ef58c6d3a5ebe0596f3531b80000f194838dab371c539ec",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:54+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_onestop.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..4729ceb7e0bf7f7b5d1ffaef7a2e76a6ed97b6b6
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_onestop.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "1064934eb2168c8212f691eb54760be2742c85deb532aa00524650bb6ed5149e",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:46+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_potec.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..f11290cfe545e590b4826a86b9cb893b1cec612a
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_potec.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "790dd61bbfa54e5cbfb7eb98bfa506425b907bbbc759eec84d66d5c798fa1171",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:18+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_provo.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..380c2d94623dd1658c291eabc832b33cd98e5f6b
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_provo.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "2529509fbf8f02fa09e5dca6798e9d6a708aeb34aac087ad6dfb66f4fff01f92",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:18:47+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_sbsat.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f7df83cee1abfcd819d2c19c2d52a9bee360072
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "a740f9f27eef77a88cad249d422e9c25a22caae3981dde6f36b3deba7b71982d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA A100 80GB PCIe",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T14:29:10+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_story_sbsat.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..cacec21f9e2cac0da3ce15e7d0ae813cab89d9b6
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_story_sbsat.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "e011a021b965bd96556883e7444eab60353703afa90bf1719a5032921ad03eed",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:46:51+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bd562c7881512d8ef380fbe23dfd7485b9002b1
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8d51f4a945f4f3c400fb3eaa143244dd629a04027e6e5279eb1e8349cbd59f11",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:14+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_szewczyk_n400.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..0bc81ac9c5896681749923f7e94f097cac047708
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_szewczyk_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "db4a955f8c1098f402d1737515ea5d1ac1c5673f81575b556f4d67210b91e10d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:29+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..9dd78e8755a300bb8d96405642e08d2505b0190b
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "5fee55f8ad58a27001fa1c3eb9b1feb596d6916f2e8a4ff4801730f6fb4e40ba",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:50+00:00"
+  }
+}
diff --git a/manifests/gpt2_next_word_renyi_entropy_a0.5_zuco1.json b/manifests/gpt2_next_word_renyi_entropy_a0.5_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e1ebbb604e6cc61ba263feb0aaedc88fff8461d
--- /dev/null
+++ b/manifests/gpt2_next_word_renyi_entropy_a0.5_zuco1.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "next_word_renyi_entropy",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "767753ad56b26dc97ca0c925a6d0e3c366f5896062b94aefab2cfaa4e1a05586",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:23+00:00"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_celer.json b/manifests/gpt2_responsivity_N512_s0_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..cdf7bd99babc1062dfde5eb79d25c41138d313be
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_celer.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b43011c8d4959e63fde3e7a4ebd439547e3801223ead1aac91f88fd79cbd6106",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:22:48+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_celer.pkl",
+    "source_sha256": "5969ca3adb06f115405faeda6aabdadb05f09a9d38eaa2cdb43b89dff4aef229",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_copco.json b/manifests/gpt2_responsivity_N512_s0_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d5db29a54c0cdc1775cf6ee2e60c710ad256880
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_copco.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e480a80beeb548c919caad4273e2e8ee61d20d5f8edcfeed6ffc47cb19276e79",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:31:45+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_copco.pkl",
+    "source_sha256": "f795a789da38b595e2ea5b69124cfd448ba62be6976b6b196b3d80dccfd0376c",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_dundee.json b/manifests/gpt2_responsivity_N512_s0_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4da1a8af2acf4658605d1e320b42fb3694b4dd3
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_dundee.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9631ca01be6209ac146704a67175e71b0fd91ad6ce07e43d18f993f7a47f320c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:39:08+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_dundee.pkl",
+    "source_sha256": "a2e99f2e72905cb3c462ccdba8d5859337fab9bcd40dc1c263ebec17c329c6a3",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_geco.json b/manifests/gpt2_responsivity_N512_s0_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..def6cf1bc10fa51815cf13272e43396449b69fdf
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_geco.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "dd01bbf2d34ad5bd61c1e123cefadc7ca1289fbcc7b5bd1fb2dcf45ac3e383bf",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:54:54+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_geco.pkl",
+    "source_sha256": "fea8b4cd36561bab5d6233297156e705f5bedff7f5fcba6f18cb79a590fd0497",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_meco_de.json b/manifests/gpt2_responsivity_N512_s0_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..2efffb1e49e5e59803cc25c8ddd5febbf51cd0ea
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_meco_de.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d7ec36035dbe54358d0647b878486f71c111902ae21efe137e95f83901932625",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:57:11+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_meco_de.pkl",
+    "source_sha256": "29b9b476899fdbb8e9680c69673d45b5e5cbc44bf4024739f4b3dde523de7440",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_meco_ee.json b/manifests/gpt2_responsivity_N512_s0_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab8fe7671ad2324f946dd8dc7d0097005352d0cc
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_meco_ee.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "96bef7386957da6bcab65ae7959e74b49a6a24c8943ee40a5c3246c28eedbf54",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:59:23+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_meco_ee.pkl",
+    "source_sha256": "77ed8a16787677768f7f6edc8dcc188399967019b81901d1db380c1de341d415",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_meco_fi.json b/manifests/gpt2_responsivity_N512_s0_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..37028ff1558b7e0adb68b3960e8c13009bbeaa0d
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_meco_fi.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2819d42573b90b4c1a085d83feb5bd3b9af0010cf0c05775bbe49d5eb912618f",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:01:34+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_meco_fi.pkl",
+    "source_sha256": "14db59e74d17a5d82db0ed1ef7dfd2ab424c96e2a9a7b32362ad167b6f2b1bd0",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_meco_gr.json b/manifests/gpt2_responsivity_N512_s0_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..d70e208cf7d0f9a68a5f45b79928ee325384f5a5
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_meco_gr.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "848c4d785d49311e2ea34b3d76f0971e8835afb63dfac96d78db8a9c26c16367",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:03:51+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_meco_gr.pkl",
+    "source_sha256": "b7e9601598ee40dc26517733435f9d92e8a6d480647bb1d86bed3de0cdb5d604",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_meco_he.json b/manifests/gpt2_responsivity_N512_s0_meco_he.json
new file mode 100644
index 0000000000000000000000000000000000000000..37768360c4ae06a658721dd85371b8634c33a56a
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_meco_he.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "he",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ed49348beb7e184f5d9baea8291cbdc0e8f6ece7dd575d993488213cd37826e5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7572559cda73298b7bd3cd64f31a64c20ba8d635bdd7f43fceda025590824989",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:05:23+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_meco_he.pkl",
+    "source_sha256": "0ebd02787f19c6cc78a03d8cfa6fd5155a77ed5135cf34b7a4424692ad79fdfd",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_meco_it.json b/manifests/gpt2_responsivity_N512_s0_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..9575c2545e2835bde469ab8e43253dc587b67a4a
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_meco_it.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b82c4fa78fc5391317b42676d8dc248fb3827ed93c34b3cb14ccbf6a7a39757c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:07:28+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_meco_it.pkl",
+    "source_sha256": "fc4bd7755304cb4b14ea2cc754f86fd87048b35100da2cf3c7c347767695e590",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_meco_ko.json b/manifests/gpt2_responsivity_N512_s0_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..f6bf01a63342f3890fd4120b048a787b0ee0b51c
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_meco_ko.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "469d3df1ce351e01fee14ee9a7b27badbc9c80430cf55b0bbd3a032bafb9ee61",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:09:06+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_meco_ko.pkl",
+    "source_sha256": "65c8e777358ea8422e7c05b8668ef42be454bb63f1e9a729bbff7533992c80c0",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_meco_ru.json b/manifests/gpt2_responsivity_N512_s0_meco_ru.json
new file mode 100644
index 0000000000000000000000000000000000000000..cb96f4ac8164cabd7fa9cfad919cf5a4e32e2dba
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_meco_ru.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ru",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "56ee0a7928fd1a4a985a1c9e9377c928263e569df23ce0f3e0961329e7e6462d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d79055ea06dc269c9a41f9d165a153bb1e0a27ea52692e764891148c97dce791",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:09:39+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_meco_ru.pkl",
+    "source_sha256": "c7dc6769cb9588209a6ecacc032333f9243f44783fa8898a0c9e5f20bbabcd58",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_meco_tr.json b/manifests/gpt2_responsivity_N512_s0_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..f49d633bf0ff6b308f773435d382872d03a19c53
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_meco_tr.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "998f1895fb925b151a96ea5ba1350b9a8b860cf9fce42a81ab2840bda6eb3946",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:10:56+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_meco_tr.pkl",
+    "source_sha256": "9b29a402be85305125032dd698f28761a3c1bca382aa7af0476a7f1ed80694ea",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_natural_stories.json b/manifests/gpt2_responsivity_N512_s0_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..d69e9d847f2edfff6963e0e5bf7bb8d5c48888e8
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_natural_stories.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9bc2cb90f89ff679cc5027a3d78fd3666f2fc948d6bf90923fc70003f4694034",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:12:37+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_natural_stories.pkl",
+    "source_sha256": "29ceda7fcbd4045d039a914ca26a39855602c392ab4fb97f106129613e8cc81a",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_onestop.json b/manifests/gpt2_responsivity_N512_s0_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..8dbf41421626eafd8c5c4b2289dbdf8aa7285926
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_onestop.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0baefd55d2e47d9076a2f0e4caa3cb2637a104ea5ac8653fd544cd640bfb50b3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:19:12+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_onestop.pkl",
+    "source_sha256": "7a50abb8f6194518d837ed6e2f087b0bb55f8448e99e107d6945263e3cc5258a",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_potec.json b/manifests/gpt2_responsivity_N512_s0_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..de5ba6c4469375c3f4caf8692dde9cc7b46e4679
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_potec.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6abfe1ba73ceda8d1792b5680527d5ec2da22aa8a4b18644b500fd335b91c8f7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:23:28+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_potec.pkl",
+    "source_sha256": "e1bf0994d748777b28be673b3791cfd7d2a2ca6746ee682330c0c8db8b72df04",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_provo.json b/manifests/gpt2_responsivity_N512_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..fb0ad6ae601ff896143ee97e3b478d259d7cab8b
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_provo.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2b0c3ebe08f7022cf53d2e196c6785a6b208c1505fa5d8a16487992efa5d916d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:22:49+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_provo.pkl",
+    "source_sha256": "63097c260a168edb6b35f5ab918b9c703e0c1899101ddc539d141d8d2e36ef09",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_sbsat.json b/manifests/gpt2_responsivity_N512_s0_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca1f4da340978f6bcd1de93209a02f1ef4c7fbdd
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d4d41f6e0c9c46251079d0067ebc96dc2b3fe2f7f110ae88b757288687443cfc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:55:54+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_sbsat.pkl",
+    "source_sha256": "3c31f7e90a6ba1f1846f344642809b7d6f97a9e700b602b90b09144a2c31aa48",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_story_sbsat.json b/manifests/gpt2_responsivity_N512_s0_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..25c31a456da300bcb7f3d573ece2c470ecb3f8f5
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_story_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fb875b3b71d11d54fd25558c03ab6efc22c5e77f7440a4794392353f2d3bfa7d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:24:27+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_responsivity_N512_s0_tokens_story_sbsat.pkl",
+    "source_sha256": "02af1360ca1f0eddd6f9795a14a3896b92e1119c7d9bbc3b968619bc30c75525",
+    "unit": "tokens",
+    "rule": "first",
+    "keys": {
+      "decontextualised_distance": "first",
+      "expected_decontextualised_distance": "first"
+    },
+    "finisher": null,
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_celer.json b/manifests/gpt2_responsivity_N512_s0_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..e98c445b1764b29d87b22a1b0a254a517dab922d
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_celer.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5969ca3adb06f115405faeda6aabdadb05f09a9d38eaa2cdb43b89dff4aef229",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T16:22:40+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_copco.json b/manifests/gpt2_responsivity_N512_s0_tokens_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe1b5a9a87bccdcfde0c50aabd3073a575be564e
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_copco.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f795a789da38b595e2ea5b69124cfd448ba62be6976b6b196b3d80dccfd0376c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:31:38+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_dundee.json b/manifests/gpt2_responsivity_N512_s0_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc2c9f2ea6add85ab380c12d2ca26a5ae7e1b8ef
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_dundee.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a2e99f2e72905cb3c462ccdba8d5859337fab9bcd40dc1c263ebec17c329c6a3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:38:59+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_geco.json b/manifests/gpt2_responsivity_N512_s0_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..65ed61fb6b2917cf8b5bd09b30146e5fb5dc5f0f
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_geco.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fea8b4cd36561bab5d6233297156e705f5bedff7f5fcba6f18cb79a590fd0497",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:54:47+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_meco_de.json b/manifests/gpt2_responsivity_N512_s0_tokens_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..3fa7df42b48bea56fa00bbaba0d7159fac002515
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_meco_de.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "29b9b476899fdbb8e9680c69673d45b5e5cbc44bf4024739f4b3dde523de7440",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:57:02+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_meco_ee.json b/manifests/gpt2_responsivity_N512_s0_tokens_meco_ee.json
new file mode 100644
index 0000000000000000000000000000000000000000..87eb62bffe8ef69d6ee7c9cb73f61ae7faf4981d
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_meco_ee.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "et",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "77ed8a16787677768f7f6edc8dcc188399967019b81901d1db380c1de341d415",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T09:59:13+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_meco_fi.json b/manifests/gpt2_responsivity_N512_s0_tokens_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..87190ac892c39391d44e2c037aefdd7720b6dab8
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_meco_fi.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "14db59e74d17a5d82db0ed1ef7dfd2ab424c96e2a9a7b32362ad167b6f2b1bd0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:01:24+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_meco_gr.json b/manifests/gpt2_responsivity_N512_s0_tokens_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..3f37c3de496e2caec95253780dd51f0a5893f32b
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_meco_gr.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b7e9601598ee40dc26517733435f9d92e8a6d480647bb1d86bed3de0cdb5d604",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:03:43+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_meco_he.json b/manifests/gpt2_responsivity_N512_s0_tokens_meco_he.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e263e06fc0648403892e1bd03be532ed29c2505
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_meco_he.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "he",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ed49348beb7e184f5d9baea8291cbdc0e8f6ece7dd575d993488213cd37826e5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0ebd02787f19c6cc78a03d8cfa6fd5155a77ed5135cf34b7a4424692ad79fdfd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:05:17+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_meco_it.json b/manifests/gpt2_responsivity_N512_s0_tokens_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..6cf37a41e153913f5710cbae98aefc643c1ada24
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_meco_it.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fc4bd7755304cb4b14ea2cc754f86fd87048b35100da2cf3c7c347767695e590",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:07:18+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_meco_ko.json b/manifests/gpt2_responsivity_N512_s0_tokens_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..5bbdc1fc2a299961259bd517c8621df87224a75b
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_meco_ko.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "65c8e777358ea8422e7c05b8668ef42be454bb63f1e9a729bbff7533992c80c0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:08:59+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_meco_ru.json b/manifests/gpt2_responsivity_N512_s0_tokens_meco_ru.json
new file mode 100644
index 0000000000000000000000000000000000000000..e486881da56ac1cf03815937bdd2df262d1179b3
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_meco_ru.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "ru",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "56ee0a7928fd1a4a985a1c9e9377c928263e569df23ce0f3e0961329e7e6462d",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c7dc6769cb9588209a6ecacc032333f9243f44783fa8898a0c9e5f20bbabcd58",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:09:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_meco_tr.json b/manifests/gpt2_responsivity_N512_s0_tokens_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..1fc26be37e60c5741f0320152983e2c69636e0fb
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_meco_tr.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9b29a402be85305125032dd698f28761a3c1bca382aa7af0476a7f1ed80694ea",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:10:50+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_natural_stories.json b/manifests/gpt2_responsivity_N512_s0_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..9487026b6ddca3ec0975129fa74a4c7ab84e11ea
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_natural_stories.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "29ceda7fcbd4045d039a914ca26a39855602c392ab4fb97f106129613e8cc81a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:12:31+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_onestop.json b/manifests/gpt2_responsivity_N512_s0_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..a67b2dad2a176f39d65db2eb688e0a52bcae413b
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_onestop.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7a50abb8f6194518d837ed6e2f087b0bb55f8448e99e107d6945263e3cc5258a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:19:05+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_potec.json b/manifests/gpt2_responsivity_N512_s0_tokens_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..ffd9db77aa13966ca2f3a68725660497854b8b7f
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_potec.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e1bf0994d748777b28be673b3791cfd7d2a2ca6746ee682330c0c8db8b72df04",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:23:18+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_provo.json b/manifests/gpt2_responsivity_N512_s0_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..1cf3c7e0820eec80e5cd28f4e05ff88c53aebd72
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_provo.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "63097c260a168edb6b35f5ab918b9c703e0c1899101ddc539d141d8d2e36ef09",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:22:41+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_sbsat.json b/manifests/gpt2_responsivity_N512_s0_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..7a383077e9246d31a329bb4050419ecf74973e61
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_sbsat.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3c31f7e90a6ba1f1846f344642809b7d6f97a9e700b602b90b09144a2c31aa48",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:55:54+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_responsivity_N512_s0_tokens_story_sbsat.json b/manifests/gpt2_responsivity_N512_s0_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..4272c311f741ba89e24b8f439d40734ec0cfd6d7
--- /dev/null
+++ b/manifests/gpt2_responsivity_N512_s0_tokens_story_sbsat.json
@@ -0,0 +1,90 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "responsivity",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 512,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "decontextualised_distance": "dimensionless",
+      "expected_decontextualised_distance": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "02af1360ca1f0eddd6f9795a14a3896b92e1119c7d9bbc3b968619bc30c75525",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:24:18+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride // 2"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples alternatives per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_contextual_N50_s0_natural_stories.json b/manifests/gpt2_simadj_surprisal_contextual_N50_s0_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac8fd41b1cb0d9ad7f93a24b3bff27170e4d0740
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_contextual_N50_s0_natural_stories.json
@@ -0,0 +1,95 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_contextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 50,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless",
+      "similarity_samples_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1964b1ff32d4de8b518c7dfdc6b3923c642e2c52e2efc902d98c24db1f752ac2",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:31:22+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_contextual_N50_s0_tokens_natural_stories.pkl",
+    "source_sha256": "b022a2f8b356cb2ab1caab6d7250ac394ce610a337239697b967e55f3adc9efe",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack",
+      "similarity_samples_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_contextual_N50_s0_provo.json b/manifests/gpt2_simadj_surprisal_contextual_N50_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..36d7640369916aa1538c7b7eadd0bc253808357a
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_contextual_N50_s0_provo.json
@@ -0,0 +1,95 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_contextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 50,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless",
+      "similarity_samples_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3815e37c3966a4124c8a4347684fe0b81810e96d77cb357e654fa19808595291",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:26:49+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_contextual_N50_s0_tokens_provo.pkl",
+    "source_sha256": "65ef7fc091d2cde50f32927b0f406107238a8efc4f4db598e8bfaad79e130fdc",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack",
+      "similarity_samples_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_contextual_N50_s0_tokens_natural_stories.json b/manifests/gpt2_simadj_surprisal_contextual_N50_s0_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..80f316d4482876ee62022c1a2ecb09a10c47d39f
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_contextual_N50_s0_tokens_natural_stories.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_contextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 50,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless",
+      "similarity_samples_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b022a2f8b356cb2ab1caab6d7250ac394ce610a337239697b967e55f3adc9efe",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:31:15+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_contextual_N50_s0_tokens_provo.json b/manifests/gpt2_simadj_surprisal_contextual_N50_s0_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..3352eae87d18a128adeecb08968b897510feb61e
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_contextual_N50_s0_tokens_provo.json
@@ -0,0 +1,92 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_contextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 50,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless",
+      "similarity_samples_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "65ef7fc091d2cde50f32927b0f406107238a8efc4f4db598e8bfaad79e130fdc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:26:41+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": 0,
+    "seeding": "reseeded with the seed at every stimulus; one draw of n_samples per scored position, in text order"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_celer.json b/manifests/gpt2_simadj_surprisal_noncontextual_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..910f70f4ff6797429669033018808128bff8954b
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_celer.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4372fb2a95a0c8f76c858485336a39f5c0712b21560b3008e1031e82dc5ed3fd",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:48:12+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_noncontextual_tokens_celer.pkl",
+    "source_sha256": "f02386168e13b83335d09bfacbff3fe577139dfd6b6b7ca1cac810669b03d55c",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_dundee.json b/manifests/gpt2_simadj_surprisal_noncontextual_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..90865719b081cee805c3a2430f5dce8ee367e876
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_dundee.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "78b72ab87e79a5f0b309e8eefbe631b04fe67aca64b5e7c3aa796c0952699b59",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:30:07+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_noncontextual_tokens_dundee.pkl",
+    "source_sha256": "f577423eb1f031a4b927074ca633c8c588b7d9422676cbfa57ae3ee7af8630e5",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_geco.json b/manifests/gpt2_simadj_surprisal_noncontextual_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..47ac4a57444efc2927d5e2c4b08c4f4c90941a4a
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_geco.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "096a2b19808d60e302dc419c88c9ac075d361de8284645b7f5ed9bee09933fb8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:37:29+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_noncontextual_tokens_geco.pkl",
+    "source_sha256": "6732daae92a200020dd54a7e6b12a421b2308db17dfa9805c65e584cdac67b55",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_meco_l2_w2.json b/manifests/gpt2_simadj_surprisal_noncontextual_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..a238c04f98aca573e64b0bf231dcc6e254dbea0b
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_meco_l2_w2.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5cfc6bd63127bdae087d0cc3aa84c79b941423d723816877b5bfea21237b8ffd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:32:44+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_noncontextual_tokens_meco_l2_w2.pkl",
+    "source_sha256": "45c4742136301fece802fb446a5026f4ad2ce7a2ec9c7b7f83391d60cbc631a0",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_natural_stories.json b/manifests/gpt2_simadj_surprisal_noncontextual_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..d8fbc2abc93665d13330c4b5d3b801bfc1d710e7
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_natural_stories.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d499b80108b2a96073228442e3b9e027ee1b28db04216594d7a7f0fb1c490c5b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:33:02+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_noncontextual_tokens_natural_stories.pkl",
+    "source_sha256": "461ecb59dc03a707e406c6773d16231aa059b3214eda737d542958acdaccf7d3",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_onestop.json b/manifests/gpt2_simadj_surprisal_noncontextual_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..6bde051e89088ef6b10044d22ffbdb87ae9f9c97
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_onestop.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "190ffd74b5f791347bbb7aa3117332d9f39b916828b19566286f452ae64a9dab",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:38:22+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_noncontextual_tokens_onestop.pkl",
+    "source_sha256": "52e47c39f58829349f520cc38c6a14a81a66342cb6e23f66e855606b4c145afb",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_provo.json b/manifests/gpt2_simadj_surprisal_noncontextual_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce0385ee6b43a96cc938fcf7c951841ceda84279
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_provo.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7d3097786d5185cd1b782e32218144295174a7f18e995ca1d2b4062b91f6313c",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:34:54+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_noncontextual_tokens_provo.pkl",
+    "source_sha256": "f14027327297c071c837437c881f9c127c7afadca1545e5011f95827b981000a",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_sbsat.json b/manifests/gpt2_simadj_surprisal_noncontextual_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab8a637db5f8ae092f821de5db07a953f2ad3bd6
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_sbsat.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "187ee2687e6129af6d5602a253da21b52194091aa69cf62e1766bf8bfa82f1f9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:55:58+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_noncontextual_tokens_sbsat.pkl",
+    "source_sha256": "9140ab14b03eacda6c5f61f74d2a6ddc8f2b063d8767c5392a164a84cbfecb76",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_story_sbsat.json b/manifests/gpt2_simadj_surprisal_noncontextual_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..a91d40033bcb7d93e2d90f0b1836c646cc1cb13e
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_story_sbsat.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a7fee2d6cb616c8efebf13be19e4ebd7ee574c5c7e4ed287ee302cf71c003094",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:36:32+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_noncontextual_tokens_story_sbsat.pkl",
+    "source_sha256": "5afa2a228ce22e8ada3c9674594dd80d2295babf3f1965ecf71898bdc3ed8f1b",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_tokens_celer.json b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ff3712ed176f804a378e1c40ef8e4ad5d5c775a
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_celer.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "dbcf2cdf631bf660e872ed8c79e7c7e87706431d3d2a5bc84ff2db3c9ed73b26",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f02386168e13b83335d09bfacbff3fe577139dfd6b6b7ca1cac810669b03d55c",
+  "environment": {
+    "code_commit": "1a79fb26aaebd3973e966aecc77411ff383c30ef",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T17:47:54+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": null,
+    "seeding": "exact, no draws"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_tokens_dundee.json b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..03af840f388b2d5ebe182c81b98b5e3a914bc3c2
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_dundee.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f577423eb1f031a4b927074ca633c8c588b7d9422676cbfa57ae3ee7af8630e5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:29:56+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": null,
+    "seeding": "exact, no draws"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_tokens_geco.json b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..1dec8f0640d5d3fe07df62bff56f917e055762b6
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_geco.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6732daae92a200020dd54a7e6b12a421b2308db17dfa9805c65e584cdac67b55",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:37:18+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": null,
+    "seeding": "exact, no draws"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_tokens_meco_l2_w2.json b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..d289ce77c01892e478f7bbec240f7c6bf35b61c4
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_meco_l2_w2.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "45c4742136301fece802fb446a5026f4ad2ce7a2ec9c7b7f83391d60cbc631a0",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:32:36+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": null,
+    "seeding": "exact, no draws"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_tokens_natural_stories.json b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..bff3c99bcfdbcfcd51f2271bf9be9f29f8cdb450
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_natural_stories.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "461ecb59dc03a707e406c6773d16231aa059b3214eda737d542958acdaccf7d3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:32:54+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": null,
+    "seeding": "exact, no draws"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_tokens_onestop.json b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..b16b99c7493bbe0db43f6aca08fe412a58ed52c4
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_onestop.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "52e47c39f58829349f520cc38c6a14a81a66342cb6e23f66e855606b4c145afb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:38:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": null,
+    "seeding": "exact, no draws"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_tokens_provo.json b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..566651f9cad23c4d76febbb609a4db4b5f5f457f
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_provo.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f14027327297c071c837437c881f9c127c7afadca1545e5011f95827b981000a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:34:47+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": null,
+    "seeding": "exact, no draws"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_tokens_sbsat.json b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ee755c228df57c23e40b6cdfabd84ef76f29423
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9140ab14b03eacda6c5f61f74d2a6ddc8f2b063d8767c5392a164a84cbfecb76",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:55:57+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": null,
+    "seeding": "exact, no draws"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_tokens_story_sbsat.json b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a048a64a743548a627f9a501c628c9b5b5cff42
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_story_sbsat.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5afa2a228ce22e8ada3c9674594dd80d2295babf3f1965ecf71898bdc3ed8f1b",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:36:25+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": null,
+    "seeding": "exact, no draws"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_tokens_zuco1.json b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..c94ab23d03bde8c86a9c8573e9b5c78049585bab
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_tokens_zuco1.json
@@ -0,0 +1,91 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "19ef6a919463be4324bd195bcd2ade7b8e3814db2b6fe126b13efde7b964ccdf",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:39:55+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512,
+    "fresh_from": "start + stride"
+  },
+  "sampling": {
+    "positions": "all",
+    "seed": null,
+    "seeding": "exact, no draws"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_noncontextual_zuco1.json b/manifests/gpt2_simadj_surprisal_noncontextual_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..884a163cbab1c9590a264071b6e62cf8abd7e3f8
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_noncontextual_zuco1.json
@@ -0,0 +1,93 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_noncontextual",
+    "storage_class": "positional",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "similarity_adjusted_surprisal": "nats",
+      "next_word_information_value": "dimensionless",
+      "similarity_adjusted_surprisal_first": "nats",
+      "next_word_information_value_first": "dimensionless",
+      "expected_similarity_tokens": "dimensionless"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4be38f4370928dc5558ccfb7a0e6c4abdc75349b4f1e7a2e03db8c4e973403b7",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:40:01+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_simadj_surprisal_noncontextual_tokens_zuco1.pkl",
+    "source_sha256": "19ef6a919463be4324bd195bcd2ade7b8e3814db2b6fe126b13efde7b964ccdf",
+    "unit": "tokens",
+    "rule": "sum",
+    "keys": {
+      "expected_similarity_tokens": "stack"
+    },
+    "finisher": "similarity_unit_values",
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_word_N50_M10_s0_natural_stories.json b/manifests/gpt2_simadj_surprisal_word_N50_M10_s0_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..89f7881f5f33c55489e1699f54e1c6fcc8b89817
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_word_N50_M10_s0_natural_stories.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_word",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 50,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "fcea94ee494edc31d3663fd26f76a63124c34980dcc181cd55d57d2ffd40b0c8",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:59:02+00:00"
+  }
+}
diff --git a/manifests/gpt2_simadj_surprisal_word_N50_M10_s0_provo.json b/manifests/gpt2_simadj_surprisal_word_N50_M10_s0_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..19d76cffaf8a043ee349596220a24a495e35d3e4
--- /dev/null
+++ b/manifests/gpt2_simadj_surprisal_word_N50_M10_s0_provo.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "similarity_surprisal_word",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 50,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": 0,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f0190ef547f3f9789e78a2ca3496ce2bf6806ac2f12bae5617019c30c6d1cbfc",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T10:43:11+00:00"
+  }
+}
diff --git a/manifests/gpt2_surprisal_celer.json b/manifests/gpt2_surprisal_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4e5e2b0a987258f85929445fe7c480a92f7e41b
--- /dev/null
+++ b/manifests/gpt2_surprisal_celer.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "997da786c1667776e7c5705c1c93e9afb59a8c6b5bcbf1e8d7e05767c4497e56",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:02:03+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_celer.pkl",
+    "source_sha256": "a811f2178d9532b39d0a20ea71e8170b97b84ac24181b4f5b5606f18930d3593",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/gpt2_surprisal_dundee.json b/manifests/gpt2_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..d9338261a27ce3fb9b39ce494443d3184e0dc5c7
--- /dev/null
+++ b/manifests/gpt2_surprisal_dundee.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "88d241a6c32ea79b26b382272247ecce49638c9994885602239690d8b3ecc883",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:45+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_dundee.pkl",
+    "source_sha256": "b3ef6330ee3806def1d7ca81baa9202d34efefd319e4172a9be42a949e485816",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_federmeier2007_n400.json b/manifests/gpt2_surprisal_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..b84ffb6d6febfacf1622b27431263f1fb180d285
--- /dev/null
+++ b/manifests/gpt2_surprisal_federmeier2007_n400.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a9bb407fc4ca6e9bb1b67b62219deb032d76306f1a78489a4e228038ae2e74b3",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:29+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_federmeier2007_n400.pkl",
+    "source_sha256": "117860a4e3d86997dbc6ee7d6ba8fb3bb0a5b62ae006874f9386d7fd809b524e",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_geco.json b/manifests/gpt2_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..360edc6f6dec644451b309f063e5ddb9be722c74
--- /dev/null
+++ b/manifests/gpt2_surprisal_geco.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7fda4de844e854df1f5a526089eb5809e7f7149267c122468c0fd93e1fca3fae",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:47:16+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_geco.pkl",
+    "source_sha256": "e3eaab829deea52fa9aac35554e8aa6c15a338072e5bb06bc5308c03c62f25a6",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_meco_l2_w2.json b/manifests/gpt2_surprisal_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..eddff332dd0448d07e4b8b6f37baeffe3ae94e2c
--- /dev/null
+++ b/manifests/gpt2_surprisal_meco_l2_w2.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a92823d56f199f4419021a1e1cb21fd5bad3635bd2ee0aa25bd4763ce4834a07",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:04+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_meco_l2_w2.pkl",
+    "source_sha256": "7014ee091eac28bbc8996d084ce1cbb9a09a4d1ab5d3801c429e34840fe77514",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_natural_stories.json b/manifests/gpt2_surprisal_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..63f4250b211aa9435b7163edca4ac321a1bde539
--- /dev/null
+++ b/manifests/gpt2_surprisal_natural_stories.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6371c337290dba286c0c3c959659183b2b0e872449aed849ccc2c57e35ae19af",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:05+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_natural_stories.pkl",
+    "source_sha256": "0274e2798744326760c41aab775d982c8127062ca03f69d05bbb8a6d642184fc",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_natural_stories_maze.json b/manifests/gpt2_surprisal_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..c311879816c7e48874543937183c6569d4b35b44
--- /dev/null
+++ b/manifests/gpt2_surprisal_natural_stories_maze.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8ebd4087eca298db9a5d38e4073851600814f6ca75e2b174f59fe086d71594ef",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:15+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_natural_stories_maze.pkl",
+    "source_sha256": "95271fbc7962bcab61f1ae19e1e062693e76c94eb8ebe0ea2de9ce6e407a53b5",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_onestop.json b/manifests/gpt2_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..368481183e45fb67cf569147d47cfa3c9fb23186
--- /dev/null
+++ b/manifests/gpt2_surprisal_onestop.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "21327245e3520be3c800c54a7e80ba5cb187d1df70b3bf83129866be549890a0",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:47+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_onestop.pkl",
+    "source_sha256": "77604de7c7ed75bd9ed6663c9eef7f63740502e4183c72296869eb0fe6d67645",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_provo.json b/manifests/gpt2_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..18daedd4d548510c5ccf9895c1042a02cc5c1065
--- /dev/null
+++ b/manifests/gpt2_surprisal_provo.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "04bcb2ceaf7bb706c657cb6eea760bd6d5c1e79f44177525e9e1753989cbcbcb",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:27+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_provo.pkl",
+    "source_sha256": "a883e2f14c192c50ed1888c4d436bfd2f78ddbc90fc8f547e0680e6190b87483",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_sbsat.json b/manifests/gpt2_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..87772e96117147c6fe160ab7ea30c6f8503451e8
--- /dev/null
+++ b/manifests/gpt2_surprisal_sbsat.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e7f3dbd328df1fc3df245aedd2409bc419c6a62953c8ac1bd12bec62cfa0346a",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:26:07+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_sbsat.pkl",
+    "source_sha256": "84ecc73dd338dd467b290ee562725a4a8a83ef958478e509acb27ce24f5c7351",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/gpt2_surprisal_story_sbsat.json b/manifests/gpt2_surprisal_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e6c11b7a2551e34db8ef7ba270d38a8283021d0
--- /dev/null
+++ b/manifests/gpt2_surprisal_story_sbsat.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cace81e4ff17b88293ae5f423e4f2a1a18cfede1ab8e4f6c58a266accc045dc1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:12+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_story_sbsat.pkl",
+    "source_sha256": "669cd56d94e109d6fb630c0241a3e59b779a739041f43a6e8d517ad738ee5393",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_szewczyk_federmeier2022_n400.json b/manifests/gpt2_surprisal_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b3e198f6cd3e7e58baf9c5bcbd4481c07e2f76f
--- /dev/null
+++ b/manifests/gpt2_surprisal_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2d488fcbd429b16f20d4a674346bff0c50f8ed2d89e32915b5cf83dc08c7969e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:16+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_szewczyk_federmeier2022_n400.pkl",
+    "source_sha256": "0f4690bdefc6768ad89f0735b0c088f31ac0e22c877326b371cc005368844a23",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_szewczyk_n400.json b/manifests/gpt2_surprisal_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..721972676947f7853b6576669479e5446e815719
--- /dev/null
+++ b/manifests/gpt2_surprisal_szewczyk_n400.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e2d79f0ae7b7995f5a33e68977f1c3dfa5aebaa18aa031ca072d412c69f1e7be",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:38+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_szewczyk_n400.pkl",
+    "source_sha256": "8341b006a2b372fec37d7a8d2a1ce7e71761cdbabc706adf1d68c16c472b4047",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_celer.json b/manifests/gpt2_surprisal_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..86fbbfd8b091145987911d0ec6b543c511333f51
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_celer.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a811f2178d9532b39d0a20ea71e8170b97b84ac24181b4f5b5606f18930d3593",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:01:45+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_dundee.json b/manifests/gpt2_surprisal_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..09646867a9b641be5ae5a0f447917e4bd042de7f
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_dundee.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b3ef6330ee3806def1d7ca81baa9202d34efefd319e4172a9be42a949e485816",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:44+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_federmeier2007_n400.json b/manifests/gpt2_surprisal_tokens_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..932d01c39d287561ee244faecbc5a2abe47b971c
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_federmeier2007_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "117860a4e3d86997dbc6ee7d6ba8fb3bb0a5b62ae006874f9386d7fd809b524e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:20+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_geco.json b/manifests/gpt2_surprisal_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d95ed39b2c063ca67d36912b9ba89b6c0992de4
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_geco.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e3eaab829deea52fa9aac35554e8aa6c15a338072e5bb06bc5308c03c62f25a6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:47:07+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_meco_l2_w2.json b/manifests/gpt2_surprisal_tokens_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..34142f6bd761d809d08e8fd5f17b11037668a4af
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_meco_l2_w2.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7014ee091eac28bbc8996d084ce1cbb9a09a4d1ab5d3801c429e34840fe77514",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:58+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_natural_stories.json b/manifests/gpt2_surprisal_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..68a448bb82e2efa7eaebb78b841468cd4a464a9f
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_natural_stories.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0274e2798744326760c41aab775d982c8127062ca03f69d05bbb8a6d642184fc",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:05+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_natural_stories_maze.json b/manifests/gpt2_surprisal_tokens_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..3fca49d592b1cc2e5d2825638196aa692f0d1886
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_natural_stories_maze.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "95271fbc7962bcab61f1ae19e1e062693e76c94eb8ebe0ea2de9ce6e407a53b5",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:14+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_onestop.json b/manifests/gpt2_surprisal_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..e566a215645aeca63c4a82ccee0533eca7ab296f
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_onestop.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "77604de7c7ed75bd9ed6663c9eef7f63740502e4183c72296869eb0fe6d67645",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:45+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_provo.json b/manifests/gpt2_surprisal_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..aad71474494cfbd02b639c5f397f1db8db65a08d
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_provo.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a883e2f14c192c50ed1888c4d436bfd2f78ddbc90fc8f547e0680e6190b87483",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:26+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_sbsat.json b/manifests/gpt2_surprisal_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff4e77dac9596a5f74f93c5635bef799d4a8115d
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_sbsat.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "84ecc73dd338dd467b290ee562725a4a8a83ef958478e509acb27ce24f5c7351",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:26:00+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_story_sbsat.json b/manifests/gpt2_surprisal_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..f9abd8281628f21c09a15be01e897fc2a4d3b71e
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_story_sbsat.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "669cd56d94e109d6fb630c0241a3e59b779a739041f43a6e8d517ad738ee5393",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:12+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_szewczyk_federmeier2022_n400.json b/manifests/gpt2_surprisal_tokens_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..d9714577f4c50ca5056f1e90d1a7c044b85d9b1b
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0f4690bdefc6768ad89f0735b0c088f31ac0e22c877326b371cc005368844a23",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:15+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_szewczyk_n400.json b/manifests/gpt2_surprisal_tokens_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5196111e4b1b0aeae22a71f164fbfc61da7768b
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_szewczyk_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8341b006a2b372fec37d7a8d2a1ce7e71761cdbabc706adf1d68c16c472b4047",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:37+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_wlotko_federmeier2012_n400.json b/manifests/gpt2_surprisal_tokens_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..98d694ab85a07058da774b3156d93a3e9b42204a
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_wlotko_federmeier2012_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "6441c612c5ff41065a98e6aab686483927420904fe0fbadfb7d45aabedee3382",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:13+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_tokens_zuco1.json b/manifests/gpt2_surprisal_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..218737af339b82d59205f3b2eb99eddda16c00ef
--- /dev/null
+++ b/manifests/gpt2_surprisal_tokens_zuco1.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b798ded703b32c5c28a5af5e76a865a5a7e0096f83671879982680387451fb84",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:43+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_surprisal_wlotko_federmeier2012_n400.json b/manifests/gpt2_surprisal_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..71b27a5b45832536050f3b3892e0de4ea5740350
--- /dev/null
+++ b/manifests/gpt2_surprisal_wlotko_federmeier2012_n400.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b2473315781abc133be4b6f6257f54ef732a0964f76ad3da4d826ecc8bd97490",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:14+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_wlotko_federmeier2012_n400.pkl",
+    "source_sha256": "6441c612c5ff41065a98e6aab686483927420904fe0fbadfb7d45aabedee3382",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_surprisal_zuco1.json b/manifests/gpt2_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..e53a39ae9da3100e01bd6af3746c2f4e4d5143a2
--- /dev/null
+++ b/manifests/gpt2_surprisal_zuco1.json
@@ -0,0 +1,84 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "surprisal",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "log_p_observed": "nats",
+      "surprisal": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cd9cf9d11693dc39ac88d0e8889265c5c7e7030ef0f7d11960f0bd2b782753a1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:43+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_surprisal_tokens_zuco1.pkl",
+    "source_sha256": "b798ded703b32c5c28a5af5e76a865a5a7e0096f83671879982680387451fb84",
+    "unit": "tokens",
+    "rule": "sum",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_celer.json b/manifests/gpt2_token_renyi_entropy_a0.5_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..de17d06375c75b108d81b38c2df3b0f074d60d2e
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_celer.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "9fa82fe9b5829875c59229eb86517da123420f700da2e4ea1368030112bcb7f5",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:17:31+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_celer.pkl",
+    "source_sha256": "2b2359589190910148bb05c8d6dc781b27d1cbc0eb9963a0d4979f2b91fed540",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_copco.json b/manifests/gpt2_token_renyi_entropy_a0.5_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae1e871268fcf6afe2ca3bc74e0dd1d9d585da6b
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_copco.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3cb8f956d092608dc8a7c2f60482dbe0fc74b0c4f0964a06474eb5483482bee9",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:19+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_copco.pkl",
+    "source_sha256": "061c4f9d0575c0282a478a2abdeb7daf6afc1ecf02a8466f6e30d93a554e2c7a",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_dundee.json b/manifests/gpt2_token_renyi_entropy_a0.5_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..5945a8ed0d35645fb63cb1a69de0c52e1b8c80c2
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_dundee.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7e4ad431d09360c3ef77b5ab59cda2c89be8a4114c7b864995cbf59bb463395a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:19+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_dundee.pkl",
+    "source_sha256": "903c6484d433eb558ab71cf6a44d5cb5df573932eab90f8d9d568d07908a9a5c",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_federmeier2007_n400.json b/manifests/gpt2_token_renyi_entropy_a0.5_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e546e94fea6d58a67eda672a29fecdfef378cb4
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_federmeier2007_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "97271a051034de93642145530c343ca5c4e333ce3990ea9c64ea7dd4447f42fb",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:29+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_federmeier2007_n400.pkl",
+    "source_sha256": "e9bbe420859be320f039f6fa7a5d20536c9e66244b3c93559a9278c4b22768c9",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_geco.json b/manifests/gpt2_token_renyi_entropy_a0.5_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..e12ebca360c775c1184ffaf214cd5b33dceae757
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_geco.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "28ad7c4cbe5b107ef066d268bdad0cf973776fbf8fa22088462874c68ff57a89",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:46:18+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_geco.pkl",
+    "source_sha256": "ec36d3febc865d7f8bd5c0e53e377e3c46f456ec555d92f1e9018d12cbb60bf4",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_meco_l2_w2.json b/manifests/gpt2_token_renyi_entropy_a0.5_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e225dee152e15aaa003a30b96bf3d53d73435ec
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_meco_l2_w2.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "181d7349efe238389b34db4eeac9ee2bf56249fe35dbf0570820df76a6a73523",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:53+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_meco_l2_w2.pkl",
+    "source_sha256": "d3349929cbb44320d79c7dc22cf3b26af62545cc5f85cff886b76732c5c52dcb",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_natural_stories.json b/manifests/gpt2_token_renyi_entropy_a0.5_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..723f2696ce0bdbfe24be62e0ffe7b37147cfba06
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_natural_stories.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "aec80cf2609bbb9509d6a3306d24ffdbfe742cb8f937017a5967806168031a7e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:31+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_natural_stories.pkl",
+    "source_sha256": "76e7e9feeecdbeaf74ad29a040d37087e01b32dc766f3b600a4b877507027bc6",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_natural_stories_maze.json b/manifests/gpt2_token_renyi_entropy_a0.5_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..201da52319d9946f781945c5f18b60fa269dbb58
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_natural_stories_maze.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "00d9d7ba703f3aefc00f6c9ec6018f923f156ffbbb5ca984359642d40c0cb233",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:45+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_natural_stories_maze.pkl",
+    "source_sha256": "aeffc9f376594f13a5243040487dceba5a09adf6acf32f561d29b9f8a108fd88",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_onestop.json b/manifests/gpt2_token_renyi_entropy_a0.5_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..d544cb9906b6bd6a89e44e5efdbdb6ebee2c33d5
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_onestop.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8af2f64e0d02bbcb7d4c4aa98ef3c417df15d764732d7d2705693b1dba68c808",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:44:32+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_onestop.pkl",
+    "source_sha256": "fa3bb545d59e33888363b4ca34164cfe303d3a736b48671af08a628288acad3d",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_provo.json b/manifests/gpt2_token_renyi_entropy_a0.5_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..34c0acd01ba8c1570e542bc97fa5a16a2bd6e9bc
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_provo.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "36e003126c010bfb721f67c6d889982a5cb534b48617259705014b4940bc50b6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:57+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_provo.pkl",
+    "source_sha256": "0f4e3bc9ebdd1c81f5057b295d20bc8b8f165c819bfca4a1b162cab6e9464974",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_sbsat.json b/manifests/gpt2_token_renyi_entropy_a0.5_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..d976ba373f1bebc40538979ee94107abac12e2db
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3cc0129d47f0c6aa172ac3437fb4138c4ae83f166f63abf8a16ffb4306f03d16",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:26:07+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_sbsat.pkl",
+    "source_sha256": "ad214d8705055192fe0e5167c96ad004bd9ff70ed9da1f48de7f4e97e6a5fb14",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_story_sbsat.json b/manifests/gpt2_token_renyi_entropy_a0.5_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c1ab29cea6c52e4bb7342f6527ff185a42b48c0
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_story_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a12aa31358ca87ffefa62e07ce27c26ae057faa3e0b61d67ce5da7f6dd6637a7",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:54+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_story_sbsat.pkl",
+    "source_sha256": "093035d24cb261250950f797a345af0ce819e3c9e3c2a8209d27628cdc22ff7f",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json b/manifests/gpt2_token_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..464862aa113822d80d20de8681fe0558c676276a
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8a4e531485013de0318a4d4ce26613634ecf2478cf55c226a27cc07335245aa1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:03+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_szewczyk_federmeier2022_n400.pkl",
+    "source_sha256": "bbb852378aeea8c7f2342a178876d9cd0edc35a6e8474558fef21cb7f64d721d",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_szewczyk_n400.json b/manifests/gpt2_token_renyi_entropy_a0.5_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..31ba30b2785bc6c9384619c42cd273e3f04e79cd
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_szewczyk_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0dea7da5e33eba06541bb1a9b80d7e70ae2974fda585712186d0758809b823b6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:26+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_szewczyk_n400.pkl",
+    "source_sha256": "805ab820350c0bc897e1f753be5db17cc2eb27a554cfb26f9ad435592289db30",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_celer.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c7c2f292a3327af4ec47597efbcf27648ce622f
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_celer.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2b2359589190910148bb05c8d6dc781b27d1cbc0eb9963a0d4979f2b91fed540",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:17:15+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_copco.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..25e787835c47ccce6de8bdc7ade0b3fe383be41a
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_copco.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "061c4f9d0575c0282a478a2abdeb7daf6afc1ecf02a8466f6e30d93a554e2c7a",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:18+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_dundee.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..a938330b00961551891e0ea32143e48d7df0f355
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_dundee.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "903c6484d433eb558ab71cf6a44d5cb5df573932eab90f8d9d568d07908a9a5c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:18+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_federmeier2007_n400.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3d7e2e0c80db4e18bf4bf499175c91745a32da7
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_federmeier2007_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e9bbe420859be320f039f6fa7a5d20536c9e66244b3c93559a9278c4b22768c9",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:28+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_geco.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e5a9bb7915495fa59e875f82d02a3bc4fcd574e
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_geco.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ec36d3febc865d7f8bd5c0e53e377e3c46f456ec555d92f1e9018d12cbb60bf4",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:46:16+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_meco_l2_w2.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..12982f3336bade2a14c8f971dc77810b139b5d6d
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_meco_l2_w2.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d3349929cbb44320d79c7dc22cf3b26af62545cc5f85cff886b76732c5c52dcb",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:45+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_natural_stories.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..2eccc0805495d67363c6acc39768271399164fc7
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_natural_stories.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "76e7e9feeecdbeaf74ad29a040d37087e01b32dc766f3b600a4b877507027bc6",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:30+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_natural_stories_maze.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ef5debfd2d01a5316195f62c6cab0d0a39b0211
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_natural_stories_maze.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "aeffc9f376594f13a5243040487dceba5a09adf6acf32f561d29b9f8a108fd88",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:44+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_onestop.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..b66d62305474ee1ca6c70a41e7be8f2bb08e859a
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_onestop.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "fa3bb545d59e33888363b4ca34164cfe303d3a736b48671af08a628288acad3d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:44:30+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_provo.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..b87cb80c763031669120fb0eafe348d91128e834
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_provo.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0f4e3bc9ebdd1c81f5057b295d20bc8b8f165c819bfca4a1b162cab6e9464974",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:56+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_sbsat.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f5f5ae367c6f1183a1ba770ce47b151b962512e
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_sbsat.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "ad214d8705055192fe0e5167c96ad004bd9ff70ed9da1f48de7f4e97e6a5fb14",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:26:00+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_story_sbsat.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f00f927040067c5f6ca9bac76b3e5d813630549
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_story_sbsat.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "093035d24cb261250950f797a345af0ce819e3c9e3c2a8209d27628cdc22ff7f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:53+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_szewczyk_federmeier2022_n400.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..33f972b1b686f5e0af528a9ffcb88f673fd0080c
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "bbb852378aeea8c7f2342a178876d9cd0edc35a6e8474558fef21cb7f64d721d",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:02+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_szewczyk_n400.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5e72446ca305ae20229f0b0a15681396032b77d
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_szewczyk_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "805ab820350c0bc897e1f753be5db17cc2eb27a554cfb26f9ad435592289db30",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:25+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_wlotko_federmeier2012_n400.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ed3c7ee424fa11ee1029114b17f789b292bcd9f
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_wlotko_federmeier2012_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "2b9f26681498b331b632e5c577f6f0bddec3b63b4075efa710797bc068589d85",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:30+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_tokens_zuco1.json b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b3ab9acd16d632615e1c9d708bf6cbba0be95b0
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_tokens_zuco1.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4016304c36c8393e2ee5062d5b1152a1ae7c07eed2f4b11abec1eba8193c988e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json b/manifests/gpt2_token_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..cbc5b503968f230aa91b22fe218a93e6766daf17
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_wlotko_federmeier2012_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "1c30e08280a6958db9ff1ef1a357cfafea9a71b4f78496d5348b0a7cfa6ad1f1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:38+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_wlotko_federmeier2012_n400.pkl",
+    "source_sha256": "2b9f26681498b331b632e5c577f6f0bddec3b63b4075efa710797bc068589d85",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a0.5_zuco1.json b/manifests/gpt2_token_renyi_entropy_a0.5_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..9173606765d37038d5f14c4793f2104405304bcd
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a0.5_zuco1.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 0.5,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha0.5": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b24700b8c24a5560e14271ebd8ee3b508a00ea3de04a8b4e0443da8bf55658ec",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:30+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a0.5_tokens_zuco1.pkl",
+    "source_sha256": "4016304c36c8393e2ee5062d5b1152a1ae7c07eed2f4b11abec1eba8193c988e",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_celer.json b/manifests/gpt2_token_renyi_entropy_a2_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..7102a56c2f44932ca82781e348c785bc3dc5f505
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_celer.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d47871cce66e46d1d7f7b82c6076d09576292c0c135be13fdbbf2c52269b9a80",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:17:29+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_celer.pkl",
+    "source_sha256": "7d96b10ff645b942323881ec280422d2649ebf1f27c6a02cefcf89923c8c565b",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_dundee.json b/manifests/gpt2_token_renyi_entropy_a2_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a600731230ab4ed6b90d8dbf9ff632728e89c46
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_dundee.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "187b3c33b0138677a30ba045f7d9b0450fb962f52a175cfc505e30cf577da183",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:21+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_dundee.pkl",
+    "source_sha256": "c3aa08f0803617cc84828feb8da486085f0779a7b41c85c417c01353870d2570",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_federmeier2007_n400.json b/manifests/gpt2_token_renyi_entropy_a2_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c4e4e2c16039f96df6f00bbb5570a867c6c8c2c
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_federmeier2007_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "16819c5c221c51b67694ccad9f3d9badc704b8b99fc04e9370cad3808761ce27",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:30+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_federmeier2007_n400.pkl",
+    "source_sha256": "491a79a185dd7d2af316213564a2516881c4b1004a39d37dbbd3569a980c277c",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_geco.json b/manifests/gpt2_token_renyi_entropy_a2_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..fcc8229a5d65542a7164afa2fc2056be4d9c1825
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_geco.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "89c71f14dd72f4f6932ff63cd22ffbff5dda9995703045ba268a0dc519c2b956",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:46:15+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_geco.pkl",
+    "source_sha256": "4f0c839e56471a97da7fc663ec32ce07ab4e2a9932da454f65173c36770ff772",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_meco_l2_w2.json b/manifests/gpt2_token_renyi_entropy_a2_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..f81c7f8a48297f24c264c764fd355668258a3561
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_meco_l2_w2.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c7c3ef94d3da8c2543e6caf4de616cfe2b74285cc05e53e6eecb6b606f2831b1",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:28+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_meco_l2_w2.pkl",
+    "source_sha256": "f0f9e989dc9e4f6b5d1d675a08d86d7cdaa97349f1cc6e9a091f08cd3e200897",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_natural_stories.json b/manifests/gpt2_token_renyi_entropy_a2_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b3b83e1fb63397b7f3611712ae980dab987562b
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_natural_stories.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "06f32492b5b0b84a8f5cdfe9a1a68c67027d9b520dfee73113b469f02cb378f0",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:34+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_natural_stories.pkl",
+    "source_sha256": "cc57d2263da020e79aeaa7143660083755cba67b57e1dcd5370a17d612166c53",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_natural_stories_maze.json b/manifests/gpt2_token_renyi_entropy_a2_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..3716955e8f851ae6e0b08913efe3de796d89e2b1
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_natural_stories_maze.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "42150774eae5958b6df9e1eaa4c509b104d9fbaf3bd820c80b503db5b9f035cc",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:48+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_natural_stories_maze.pkl",
+    "source_sha256": "649dae6cbdd526eb6d34bfc157f5ed2cd9044b0bbdea3cee0e3621bc101126a7",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_onestop.json b/manifests/gpt2_token_renyi_entropy_a2_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..5c9b6641e537163055a1feea56dc38debd37e449
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_onestop.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "48c898d08f6c1002adfc61ae7593cac9e1d609adf69d730fbeb03c0efe4c742f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:44:29+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_onestop.pkl",
+    "source_sha256": "3f275c40f91cbf2aaa11140c788c818adede26e945ef5b58dac6631c1648d2c4",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_provo.json b/manifests/gpt2_token_renyi_entropy_a2_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..a215d1e867a439259480d6a9f5c5ff4d648d0baa
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_provo.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "d5adc928ef1097fa98906138b18941353d26cb5b740684a48f50879c02385ac8",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:59+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_provo.pkl",
+    "source_sha256": "5d26afb7590ddfd07d9f85f643fc1c26cd666b63bc03c6aecfcfa505fdfc8625",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_sbsat.json b/manifests/gpt2_token_renyi_entropy_a2_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..841b8ff10e38f8dfc7da7de29656ef08b1709769
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "0380b295d7efba050b715a24b95cecfbf578b7e22a3cd2c4cf595ec6649fc5c3",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:27:54+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_sbsat.pkl",
+    "source_sha256": "88e02c51381fe5aa5388193326e7e7af775c6b6e69103513ed916faf48fe6062",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_story_sbsat.json b/manifests/gpt2_token_renyi_entropy_a2_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..115c06b3821ec3e655c38d0b6e64af48e3226593
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_story_sbsat.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "78ebc433bb22b0ab6313c2981b1622c3aa72d2f4b9debac375ae5eab05f2593e",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:56+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_story_sbsat.pkl",
+    "source_sha256": "a4ca976aec71e1be5611a798b4e8226f1b7d8054a869ae5cc6c18fc3b2319cdf",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_szewczyk_federmeier2022_n400.json b/manifests/gpt2_token_renyi_entropy_a2_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..c68ca3f1ea600c36104ed03b2c1a2ab43828c495
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "8335a2e1ec240b451ed3c74d651b13f58a73332cc4e28809b1d873645246553f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:01+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_szewczyk_federmeier2022_n400.pkl",
+    "source_sha256": "e454415f10722e5a4a094146402c33a52c784923ecf9637de84754630cfec057",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_szewczyk_n400.json b/manifests/gpt2_token_renyi_entropy_a2_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..919a118bdb44bb26215670a2ead7b1b2baca9ec3
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_szewczyk_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4a6640625e6c3e3bc4696a0688ce465210d9c23a047d2f7757c65b489b5244d4",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:31+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_szewczyk_n400.pkl",
+    "source_sha256": "a4b3245a33c48a3537058553b81acc6daafcb8e19134ed4bc44967281ff2a08b",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_celer.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_celer.json
new file mode 100644
index 0000000000000000000000000000000000000000..2812b7bcff5fc8ab3fc7767c9d3ab3db21bcee32
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_celer.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7ec1058e5169feed00e160622998671316bba0f47990e35c048f55e66e65fabf",
+    "n_stimuli": 28535,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "7d96b10ff645b942323881ec280422d2649ebf1f27c6a02cefcf89923c8c565b",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T03:17:11+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_dundee.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..c64eb90f9730901143b215fe99ad458cd98dae32
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_dundee.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "c3aa08f0803617cc84828feb8da486085f0779a7b41c85c417c01353870d2570",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:19+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_federmeier2007_n400.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..028afc50ceb033f228bb3f0c737155e7b5a37fb1
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_federmeier2007_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "491a79a185dd7d2af316213564a2516881c4b1004a39d37dbbd3569a980c277c",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_geco.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf53634c630c5015da26abf2f8288c2e81738f57
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_geco.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "4f0c839e56471a97da7fc663ec32ce07ab4e2a9932da454f65173c36770ff772",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:46:14+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_meco_l2_w2.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e6bca54b8bce9dc6f5b46ce735ae46f868ded09
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_meco_l2_w2.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "f0f9e989dc9e4f6b5d1d675a08d86d7cdaa97349f1cc6e9a091f08cd3e200897",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:27+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_natural_stories.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..459c01ffefcbe7c08a2234ed0e0febcbb88d7f1c
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_natural_stories.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "cc57d2263da020e79aeaa7143660083755cba67b57e1dcd5370a17d612166c53",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:28+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_natural_stories_maze.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..f66522719a92854472136124eccd88a9946c89dd
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_natural_stories_maze.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "649dae6cbdd526eb6d34bfc157f5ed2cd9044b0bbdea3cee0e3621bc101126a7",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:47+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_onestop.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b93d7328fefa70d1a64fc852093f8a48208af76
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_onestop.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "3f275c40f91cbf2aaa11140c788c818adede26e945ef5b58dac6631c1648d2c4",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:44:28+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_provo.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..023be0ed683aed3b067cbdcf38666673ef31b056
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_provo.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "5d26afb7590ddfd07d9f85f643fc1c26cd666b63bc03c6aecfcfa505fdfc8625",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:59+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_sbsat.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..d22e685a00f8ff3be8ee64b6d8ff0690018e586b
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_sbsat.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "87444aa68fe80a4b30e74a80aaa8d88876070852c140c4729945fa69118d4506",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "88e02c51381fe5aa5388193326e7e7af775c6b6e69103513ed916faf48fe6062",
+  "environment": {
+    "code_commit": "5db5c7158e68a026c204a60b2a6c3b3837961337",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T06:27:45+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_story_sbsat.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_story_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..a2875d9c85e840b6755019f2b7bc559aecb403ab
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_story_sbsat.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "story",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "a67a6402a9172eac0276b0a36d1cfbb824666522252fd3c714ad2ce1a91c5b32",
+    "n_stimuli": 22,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a4ca976aec71e1be5611a798b4e8226f1b7d8054a869ae5cc6c18fc3b2319cdf",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:41:56+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_szewczyk_federmeier2022_n400.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..b8bbf352cb02d2b92ec43aea1aa20ecea10cd15d
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e454415f10722e5a4a094146402c33a52c784923ecf9637de84754630cfec057",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:01+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_szewczyk_n400.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a562bb504e72ff784c96f64e0e5a7488870fe1e
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_szewczyk_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "a4b3245a33c48a3537058553b81acc6daafcb8e19134ed4bc44967281ff2a08b",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "Quadro RTX 6000",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:31+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_wlotko_federmeier2012_n400.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..77ab6f7ec13bb5ce811d8f694719033fb343a0fd
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_wlotko_federmeier2012_n400.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "58cfcc5f05bc91a0ec0a3693dfbb86ae070384d47de8d9b1374740390fb0c3fc",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:02+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_tokens_zuco1.json b/manifests/gpt2_token_renyi_entropy_a2_tokens_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..943391aef8093f614651412c2327586e2ea53c03
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_tokens_zuco1.json
@@ -0,0 +1,81 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "tokens",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "313d60b3f6004d48e05c363fa710da715d3da70a4280afd6fb7bd1ea74e43623",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": "cuda:0",
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:29+00:00"
+  },
+  "window": {
+    "rule": "sliding",
+    "max_positions": 1024,
+    "stride": 512
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_wlotko_federmeier2012_n400.json b/manifests/gpt2_token_renyi_entropy_a2_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..85bf386b31e40538917604ec8ba7cd78582cd0c3
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_wlotko_federmeier2012_n400.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "e702ab5210f9e07bc752947c3a02f100384914b354db9d9ec5efa5f248878e8f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:42:02+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_wlotko_federmeier2012_n400.pkl",
+    "source_sha256": "58cfcc5f05bc91a0ec0a3693dfbb86ae070384d47de8d9b1374740390fb0c3fc",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_token_renyi_entropy_a2_zuco1.json b/manifests/gpt2_token_renyi_entropy_a2_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a7351e967446025c0fa51f5d24f7b43ecb4ea6a
--- /dev/null
+++ b/manifests/gpt2_token_renyi_entropy_a2_zuco1.json
@@ -0,0 +1,83 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "token_renyi_entropy",
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 2.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "e3087b24289bd32731ed80f7be03f11aa7b58141df18e9788a9589380148c008",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "units": {
+      "token_renyi_entropy_alpha2.0": "bits"
+    }
+  },
+  "status": "completed",
+  "payload_sha256": "b3cc7a1c1583347fc363d3b0c56c3ddcee861393fe6f6279685779a3d66e348f",
+  "environment": {
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-19T18:43:30+00:00"
+  },
+  "derivation": {
+    "source": "gpt2_token_renyi_entropy_a2_tokens_zuco1.pkl",
+    "source_sha256": "313d60b3f6004d48e05c363fa710da715d3da70a4280afd6fb7bd1ea74e43623",
+    "unit": "tokens",
+    "rule": "first",
+    "code_commit": "fb4e293c972297d06b7db0697db8e2d13925986e"
+  }
+}
diff --git a/manifests/gpt2_wordsprobability_surprisal_buggy_natural_stories.json b/manifests/gpt2_wordsprobability_surprisal_buggy_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..1da83203331af06f44b4e05da99ba2845b6b075c
--- /dev/null
+++ b/manifests/gpt2_wordsprobability_surprisal_buggy_natural_stories.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal_buggy",
+    "storage_class": "positional",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "pimentel2024probability"
+  },
+  "status": "completed",
+  "payload_sha256": "5471e5521ae4c3f860d8119128b7b58156a84b7fdb26f538ede0c269f5ecf5d5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:20+00:00"
+  }
+}
diff --git a/manifests/gpt2_wordsprobability_surprisal_buggy_provo.json b/manifests/gpt2_wordsprobability_surprisal_buggy_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f238deb8ec28bea649d9191e90dc88bbd8fe9d0
--- /dev/null
+++ b/manifests/gpt2_wordsprobability_surprisal_buggy_provo.json
@@ -0,0 +1,76 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal_buggy",
+    "storage_class": "positional",
+    "layer": 2,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "study": "pimentel2024probability"
+  },
+  "status": "completed",
+  "payload_sha256": "24b6c6e22b47a86e5a89d737cec9cdb7f0f63ea24ed15aef4832e2961f15c7a6",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:30+00:00"
+  }
+}
diff --git a/manifests/gpt2_wordsprobability_surprisal_natural_stories.json b/manifests/gpt2_wordsprobability_surprisal_natural_stories.json
new file mode 100644
index 0000000000000000000000000000000000000000..1541b31b1f7ba259b43b6547eaaec28f256795f6
--- /dev/null
+++ b/manifests/gpt2_wordsprobability_surprisal_natural_stories.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
+    "n_stimuli": 10,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6cc9bcbb5f8e224ef9db7eb4cad4ba603aa5796ca8c114f86fc07ce9feea9c0a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:19:57+00:00"
+  }
+}
diff --git a/manifests/gpt2_wordsprobability_surprisal_provo.json b/manifests/gpt2_wordsprobability_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..b315b343f857c4b42ccb2f740e6fe53be4bda12f
--- /dev/null
+++ b/manifests/gpt2_wordsprobability_surprisal_provo.json
@@ -0,0 +1,77 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "wordsprobability_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "gpt2",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "gpt2",
+      "revision": null,
+      "sha256": "c5a644aeb1cf4cde835b077c20075efa570b2b51cec32626f8008b4fe580107e",
+      "files": [
+        "config.json",
+        "generation_config.json",
+        "merges.txt",
+        "model.safetensors",
+        "onnx/config.json",
+        "onnx/generation_config.json",
+        "onnx/merges.txt",
+        "onnx/special_tokens_map.json",
+        "onnx/tokenizer.json",
+        "onnx/tokenizer_config.json",
+        "onnx/vocab.json",
+        "tokenizer.json",
+        "tokenizer_config.json",
+        "vocab.json"
+      ],
+      "requested_revision": null
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "142be7b49ee259cfa325e62d31fec9655806b286f68bbb577594efb923b3a961",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:01+00:00"
+  }
+}
diff --git a/manifests/wordfreq-da_unigram_surprisal_copco.json b/manifests/wordfreq-da_unigram_surprisal_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..eeb3c3b84f86a25cd387143c9982e78bc46c3b08
--- /dev/null
+++ b/manifests/wordfreq-da_unigram_surprisal_copco.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq-da",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": "da",
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq-da",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "1cb3849b2e9354f3b927d7d58aa5c30188dda2843c868961a7d7077955d95d3a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:08:18+00:00"
+  }
+}
diff --git a/manifests/wordfreq-en_unigram_surprisal_onestop.json b/manifests/wordfreq-en_unigram_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..78136bd7d3a98dbec50daea9d7f1c69986b5e23a
--- /dev/null
+++ b/manifests/wordfreq-en_unigram_surprisal_onestop.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq-en",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": "en",
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq-en",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6094698de3ea91184937cc776e85770ca539d088cf8bfd96593fffa7e913b81d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:23:30+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_copco.json b/manifests/wordfreq_unigram_surprisal_copco.json
new file mode 100644
index 0000000000000000000000000000000000000000..487bdbec8770afb501c45dd502f8fb082aade9db
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_copco.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "da",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "f6fd7125d1cdfea9a1bf66716a070d46c5168c5e319540f19ad4fc98be8b5619",
+    "n_stimuli": 452,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "1cb3849b2e9354f3b927d7d58aa5c30188dda2843c868961a7d7077955d95d3a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:08:20+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_dundee.json b/manifests/wordfreq_unigram_surprisal_dundee.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec74e5bf60c42b2b444a20afff3b1447845757aa
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_dundee.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
+    "n_stimuli": 20,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "9eaa615d7664be2f218f8a47ab10a3b2303a8e98772f469b991d984469a91a74",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:28:33+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_federmeier2007_n400.json b/manifests/wordfreq_unigram_surprisal_federmeier2007_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..446c5b438c1d3baa5bf37052bf0eac10ed07127b
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_federmeier2007_n400.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
+    "n_stimuli": 564,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "4c319e7501b3b774eaa1bb7abbef384720a8c34bed756d1290bd40b2944df095",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:08+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_geco.json b/manifests/wordfreq_unigram_surprisal_geco.json
new file mode 100644
index 0000000000000000000000000000000000000000..06d0f7daeff736f6fd9adf626c251021004cb7f6
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_geco.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
+    "n_stimuli": 588,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "88f6a0aa6a7f961cfb736557f0faaa931110ff07cf614ee17dfcfa683e02e8d3",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:45:48+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_de.json b/manifests/wordfreq_unigram_surprisal_meco_de.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb39e1ed14d03fbbccd023c4e7368a6f3840c174
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_de.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "35097aa7c5881e1082b0e89ffeb3e9d9f172088adf3a5535d7eaa904d50b1239",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:44:37+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_fi.json b/manifests/wordfreq_unigram_surprisal_meco_fi.json
new file mode 100644
index 0000000000000000000000000000000000000000..21eb5287ae571ee6242c87320f4d2e72efd8b17b
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_fi.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "fi",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "765ebc2798ba14778d53894bdfa0bba45ee1917db8b114300bbcc505089dc058",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:29+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_gr.json b/manifests/wordfreq_unigram_surprisal_meco_gr.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee7931f92b746b49471a572915c7c4841839667d
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_gr.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "el",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ca453b3daa0b2f71984f2beed95ee06ded8caff8d6e982502c466e45b04233fd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:36+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_it.json b/manifests/wordfreq_unigram_surprisal_meco_it.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb2a883aaad509feab38ceb53731fd416d9e8a38
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_it.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "it",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "8ae08a110f9387d964ea3b8db17820bfc194866732f216d4cb0b156302185022",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:34+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_ko.json b/manifests/wordfreq_unigram_surprisal_meco_ko.json
new file mode 100644
index 0000000000000000000000000000000000000000..1e0d707871f0681f41a36846d7564728c96fb06a
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_ko.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "ko",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "753614581e2ec0f145df56bcc407566bb4f572f08da9408586cdee55dc8f78eb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:45:57+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_l1_w2_bp.json b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_bp.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec4643e63d8314e74b77abae7452bb63f0532b97
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_bp.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "pt",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "7787d00f660f0c84e6e44a4915ef1e1b89eef7391cc55d9eaedf14172ab77b0e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "d17d62deea88d57df35afad1a4bc031480b57c273a77968d875463ee2aff242a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:42+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ch_s.json b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ch_s.json
new file mode 100644
index 0000000000000000000000000000000000000000..445ce651eb77fdb0207aa3ab38bf7309108d3d1e
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ch_s.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "9f27685ea775f0843723c5aa8c5b226b7d3420ccb213afd1eb8198439dcdb6a3",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0f9722e08d3df9b43a016835adc3ad6ce9aff6906ad58c36d0c682dfae047318",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:46:09+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ch_t.json b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ch_t.json
new file mode 100644
index 0000000000000000000000000000000000000000..11aeb2301b7ae0d3b0716f381babd1857f3fd2d2
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ch_t.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "zh",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "866884887c127db9b93a7cf7e4deb154d4c2cba960aa58ede0d05493044e3877",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "18c0e863c773b0a40f6545e124179d2030d09537ff43639666ac03414e2aa93a",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:46:14+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ge_po.json b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ge_po.json
new file mode 100644
index 0000000000000000000000000000000000000000..429f75a62b4e95c9c4ba4d68a48a078e329feae7
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ge_po.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "96ffeb58ee3b6499c665737a710b7d03d1bffba213589bb09d3c3afda4def237",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "cc70ad63e3d17c69d0ce57c7962dce19b8bd3748e2e1a873be568746b42c70fb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:48+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ge_zu.json b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ge_zu.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf42c50bb4c19cdcd9867e96cf51e84ae24c9943
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ge_zu.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "c28be2cf8ee175aea13f88564e64f7c6fa95eb540c1c258d823ae1e37303669e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "298f699bc0bdf08439687684b83213ca2625a4d6394142197c71471cb7c45065",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:48+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ic.json b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ic.json
new file mode 100644
index 0000000000000000000000000000000000000000..fa5dddb19c3a651fe860da4f401c311dc86ee333
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_ic.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "is",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "33cec4d1151c03d34cb55bd029757cc06b5dcad91c5b58937ad29f54fd39a23e",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f5ae9be37963db6b8f7aae7fe8ca8cbc4894faada7ab6ef77258d252b8523fdb",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:51+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_l1_w2_tr.json b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8e6d7ba7e53ae2a8ff46c60556d4278512b0cd9
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_l1_w2_tr.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "01c5b893329cc0a9a1cb32444221352edd389042fdf157931d862342df42b862",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6eee1d7d95a61ea5c05bee46a70cdfa13165441fc6277bb993a7c01cc1c606dd",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:51+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_l2_w2.json b/manifests/wordfreq_unigram_surprisal_meco_l2_w2.json
new file mode 100644
index 0000000000000000000000000000000000000000..cfe2d3c5172d14632428a1e3453cbcb0fd56a314
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_l2_w2.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
+    "n_stimuli": 18,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f405f910b40bf9cf1dffce16445551ac711529e52b5125c3b27c3bfc5e005cef",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:45+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_meco_tr.json b/manifests/wordfreq_unigram_surprisal_meco_tr.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd98ba9b55cd8386a631bd262b5c635eb8e68d01
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_meco_tr.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "tr",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f8137cbcb33eb12c67289ad4385e74b9cd55800a175f46cc5ffb010be15e77df",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:20:48+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_natural_stories_maze.json b/manifests/wordfreq_unigram_surprisal_natural_stories_maze.json
new file mode 100644
index 0000000000000000000000000000000000000000..61f326431a0d5203d62f31ff0ccdc41d93c2351d
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_natural_stories_maze.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
+    "n_stimuli": 474,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "0a7faab50111adcc9d331de3bece2f50f7006ce02bcd4be8624facc6b5c5f531",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:10+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_onestop.json b/manifests/wordfreq_unigram_surprisal_onestop.json
new file mode 100644
index 0000000000000000000000000000000000000000..a9c1b83d742977ed82cb99b8b132a67281a59ec8
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_onestop.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
+    "n_stimuli": 339,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "6094698de3ea91184937cc776e85770ca539d088cf8bfd96593fffa7e913b81d",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:23:24+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_potec.json b/manifests/wordfreq_unigram_surprisal_potec.json
new file mode 100644
index 0000000000000000000000000000000000000000..18c501cae11e18999fb1689c2409e6147c579edf
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_potec.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "de",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "c7e39eb5101193ad025d1524c0714589c0f5ddbabdaab1ec5fc6b95796399ea6",
+    "n_stimuli": 12,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ca3e1be49c366a06a91667e9af504f54c27f7758f77fdd3dc20560f84189c429",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:22:29+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_provo.json b/manifests/wordfreq_unigram_surprisal_provo.json
new file mode 100644
index 0000000000000000000000000000000000000000..d4488761cb8b63ad7262a5e24fac149a6430c889
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_provo.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
+    "n_stimuli": 55,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "598b0ee7ae08016ba02430614264bef2144fbccc145c0820331d8b79d2f15a49",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:07+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_sbsat.json b/manifests/wordfreq_unigram_surprisal_sbsat.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f7bae0fe161d4d13b2b5fe734713bb9f60f218f
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_sbsat.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
+    "n_stimuli": 22,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "c0f95b5dcba7da8bc1e9a4771e47ee7f6af4918ea3c81dc7dd61f71a0c8402f5",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA TITAN RTX",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T11:46:58+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_szewczyk_federmeier2022_n400.json b/manifests/wordfreq_unigram_surprisal_szewczyk_federmeier2022_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..2fcf6ddbc2debdc8e18971582b05f6e9811aad1b
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_szewczyk_federmeier2022_n400.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
+    "n_stimuli": 600,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "073dc8ab302dce23d9eb49ca0f167ed8ab6f272effe327444e31480e38d0c083",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:07+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_szewczyk_n400.json b/manifests/wordfreq_unigram_surprisal_szewczyk_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6dfd2e133d0616dcd53c92fb75d802619b9acf6
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_szewczyk_n400.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
+    "n_stimuli": 1331,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "ca4c553c475932c94bacedbaae372f1e55898eac81b228925ead4bf53a35a3c9",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:24+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_wlotko_federmeier2012_n400.json b/manifests/wordfreq_unigram_surprisal_wlotko_federmeier2012_n400.json
new file mode 100644
index 0000000000000000000000000000000000000000..31108311ab8924b3427f5ddc0456b8e939b5f3f8
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_wlotko_federmeier2012_n400.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "f6f07a03f0395419171f21806efb50832e5e1d543377c0b72f4ef5d239ef5f63",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:21:07+00:00"
+  }
+}
diff --git a/manifests/wordfreq_unigram_surprisal_zuco1.json b/manifests/wordfreq_unigram_surprisal_zuco1.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3dc8f0cdcdffffef2a6be3c7bb6985410363b00
--- /dev/null
+++ b/manifests/wordfreq_unigram_surprisal_zuco1.json
@@ -0,0 +1,59 @@
+{
+  "identity": {
+    "schema_version": 1,
+    "method": "unigram_surprisal",
+    "storage_class": "unit_bound",
+    "layer": 1,
+    "options": {
+      "model_id": "wordfreq",
+      "beam_width": 5,
+      "n_samples": 64,
+      "max_tokens": 10,
+      "alpha": 1.0,
+      "seed": null,
+      "add_bos": true,
+      "dataset_language": "en",
+      "language_override": null,
+      "chunk_size": null,
+      "word_starters": [],
+      "top_p": 0.9,
+      "context_scope": "stimulus",
+      "unit": "words",
+      "window_policy": "reference",
+      "segmentation": "whitespace",
+      "model_revision": null,
+      "store_distribution": false,
+      "model_identity": null
+    },
+    "model": {
+      "source": "wordfreq",
+      "revision": "package-version"
+    },
+    "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
+    "n_stimuli": 300,
+    "source_sha256": "40fd4bc86f96809413e4945f69d1d59272b2de15ac949a4e90f80ec4e90434d4",
+    "versions": {
+      "python": "3.12.13",
+      "replica": "0.0.1",
+      "numpy": "2.4.4",
+      "torch": "2.11.0",
+      "transformers": "4.57.6",
+      "tokenizers": "0.22.2",
+      "wordfreq": "3.1.1",
+      "wordsprobability": "0.17",
+      "genlm-bytes": "0.2.0"
+    },
+    "unit_rule": "whitespace_continuation",
+    "unit_layer": "words"
+  },
+  "status": "completed",
+  "payload_sha256": "df1fb74eb4ca81e70ccfc6a524c86814bf2d54db1bf7d58b05920c521d6695b4",
+  "environment": {
+    "code_commit": "e78a32c204cc3589c964824ba60b43c1ff2bde8b",
+    "code_dirty": false,
+    "device": "NVIDIA GeForce RTX 2080 Ti",
+    "model_device": null,
+    "cuda": "13.0",
+    "computed_at": "2026-09-20T02:22:27+00:00"
+  }
+}
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_celer.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_celer.parquet
index 48e1b75da6567d9cafc973afc6e70cade30edc9e..7f9ede5f3e80044403155d5a3be53dd1ed014d53 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_celer.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f4ee156ae854da190e6bfff419563ee2fd598a195a4e1a628889d0bbd5219d99
-size 434436755
+oid sha256:bcf8ac552bb4859b34b93561322e58a824d0b858033698f62be52a5322ffd03b
+size 433284249
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_copco.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_copco.parquet
index 55eaa5549f001364f016ed340ab3ca01ee6c9a9e..dd61358c4d16c0adea0cf30704edc29608c1d4af 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_copco.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_copco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:147844cd5e9343bd70fac2865c12a71302ea4ef6ee42ed277b70198b08a4855c
-size 40038856
+oid sha256:112bef3340067749b928a43fb111071ec97b377172f39893eee53883544758a5
+size 39963082
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_dundee.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_dundee.parquet
index 7b2a8a49e665a92c5a06f00158c8313ce2a8aff2..6599c4e9daa2ff4ffcb24ae922cc161652691502 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_dundee.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:929a9c206c449f5939c6cb418483bed957458b83d68be992c2a8195b6a8c0b3d
-size 5578199
+oid sha256:50fbec32bee81f8b0a06adf93f516a9a02d6a888a8114754305f5d157cd94577
+size 61416088
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.parquet
index e44b4f13e57cfb884b79e761386647e688c48a8e..91872c949a87d1c3d67694d751469a66b77434d3 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:24392cb0628c1b7af3f5952b10bed0527afeb060838926391b480a75057ec459
-size 8509791
+oid sha256:ce764179925735d3dfed579c2253571ca1e16d8a502cdd315281c250757b9fef
+size 8512765
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_geco.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_geco.parquet
index 894fb4988ad21f8eb9942055cf4a57de6f2ecd88..e665fa9a31b77e2095cc6160f55e061d16a07ece 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_geco.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c71ec9ed5f24795bb62af9b2cd3222c519c78a7e87f2384b9ec0007ae241a678
-size 73165338
+oid sha256:8878d426ce14b64eaefb1c737d3169c81e749b43eee4d0c87329daedc4e820e9
+size 73157288
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_de.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_de.parquet
index e285fd42b832ee80602c0adfbc99f28a9099f7c7..765b8c8c06f6789f949558ba87e9dc210baf494e 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_de.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:59c1b08f18d2f16646553dc3b36d5ca065b26d274cae59c96f039f69ab629e38
-size 2952659
+oid sha256:e3d1e5f6ac18d38d26a6cbd298d61440edfe03ecf18d5e16e200b9675b6e8516
+size 2944190
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ee.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ee.parquet
index 8f1d8257382b203e679bd7093010e82e41b31192..584a9b284a32d036eb4a14b889f34a2021f9abbe 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ee.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:75da88c6e00834d0e5d7fc770baf25a62f139bcd5bb2b17fea38d52b2ab0c4a4
-size 3306655
+oid sha256:3c90e89e85c228e62772ccd93a629b8c8695a7ed6dddc9551098928fed88a572
+size 3300108
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_fi.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_fi.parquet
index fca2ef1e08a8c3452899ba4cf6578c401f79e563..401f6060a87bffbea385bf40d8e60eef9f31ce8b 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_fi.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ed0fbd3332a34a4807bff6ae8771891eed8b6c9c3e09fd9358bb146ae056db65
-size 2690077
+oid sha256:79c4ea6c73a28bef58adfa8002567ca702d00b55ee7a157812e2c8cca636da62
+size 2676997
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_gr.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_gr.parquet
index 89a52e5e579831445b17e4d0581290f8933972ed..1a2043af4b6cf6adffba6a4e8e02fdf730177c27 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_gr.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:953dfadb9cbdf6075acb8124f24cc5fa79d0963bcff08a66de5cd4be44a0f7d4
-size 3021915
+oid sha256:61fa45875da958715e2af834609136e2604942f72f092fce92cd543e6b1fcd66
+size 3022084
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_it.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_it.parquet
index 7a91da19d772c6c15737eba6ea36451e3fb8807b..a8918ec7e3c7c7e0acb06f43bf13323df350dde4 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_it.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8e8e9653dc411ba439fbd13f3433544bb72ca6a9771d952b2be0f39405333c0d
-size 2866601
+oid sha256:d8c59bfa877325d37da6af257450da7ffca5cb3409e9be27ccd1f8c391aca883
+size 2874962
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ko.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ko.parquet
index a765b6e81a0d079e724e7bbbde2d1d8cb038a4fc..a16e67ec1748677fbe7a9f74253cd081acb9a172 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ko.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_ko.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3bdd915054fecd5a2c53155f29235eebfe01662960c3cc52795920bed9a03ec1
-size 3229408
+oid sha256:da74f42f5e57ad159abc4f466bddcf27720e66348b1d1bf39cb7f3b56b3f6e9b
+size 3228827
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ba.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ba.parquet
index 5a4734d51f86e99850d3f8af30b4f2999248a341..61b65ae6c761fd6b2fba028bd2a8552be0474ff1 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ba.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ba.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e81dd8256b49c576b5d0dd4eb97000889fc7f5ce63394b4eaf861e412a4fa594
-size 2972550
+oid sha256:3a137383d13c541ae1ca4433536a81ef665a59c9c09c9e46cfda1b3f30ed7cd3
+size 2964671
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_bp.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_bp.parquet
index 411e18c93caa94096c8b16a84852398f143b4313..26fe04a2cbac3f594bd5118ff01cbc6292b3c8e3 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_bp.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_bp.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9f7ea44f9b8768f8b262301db3a9d3454a7eb984ce83bb3b731ddfb525024bf1
-size 2974398
+oid sha256:752316fc24b1b52adadb02153eef80d34247541a716c199b057b50b80acef8d1
+size 2974845
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_po.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_po.parquet
index eff2f2c6ebb79adec33c84891787af935d5f9cb5..437e58403455b3ea374acda15e92062aa02f1f32 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_po.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_po.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dcd5ed7bd36324463e00becbad3dcdc7a00d291e869e58af3fa1019b9460aefb
-size 2951879
+oid sha256:9df8e485a3f27dd817e23d42bbd3b9dea2112570a12c70e58c9810ea926f7c02
+size 2939201
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_zu.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_zu.parquet
index 45f8c68e139c0a0c06233ad816cfbbc302fb1334..5d7ff60c36fee6bcc8c6a3038f18fc268c76fe3f 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_zu.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ge_zu.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f36118f728904dd9f2578931ddc1a39f61f9951badd1536be5dd3d28eec564aa
-size 2951879
+oid sha256:26f2e5e43610a6a82b52969346e6b05e249b55f10ee8bd45093b9ca7f5d6b55b
+size 2939201
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iiith.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iiith.parquet
index 152015083140af762e67ba20bbecc7c709a0c12a..e8d2fc2d48d598aca11bf4c7126cda48f14f0edb 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iiith.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iiith.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:47b7b569ac034e41d398ebe594032f0b22d02e536e20348e42249e5cc1e15ea2
-size 453780
+oid sha256:8ddc1fc3a0d352b26b7fa81abfdb009edb7dafdbaed8bfac8f0cebe973189efc
+size 3033994
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iitk.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iitk.parquet
index 48235f7c48e58860c9c4f5bd63a97e3c01d53c4f..9ca50f474e19c71de664013b01bbdb3f44971507 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iitk.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_hi_iitk.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8035f49bfb49653917a806dc2f72351710a3b716bea3b25b1ac19815e5057bcf
-size 453765
+oid sha256:487daae704deb4c6a37369c1ece962bc8af7df1d67d78e7a324c75c8741dd013
+size 3033979
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ic.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ic.parquet
index 859ad361d58157fdff60c9cdf8f911ec279b46bd..fff8d3760cb36ea976e1a20f0812c049fae5427c 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ic.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_ic.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:03e6d7b53a478ad23072449180dcd156c803b0522c69ff453c3639da97e473b0
-size 2205083
+oid sha256:946fcaf1e1de4f0aee6d45e70dfbd6fc4472f1d0205cca55ddb16c3a22f76775
+size 4300129
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_tr.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_tr.parquet
index d1f6e4a19d257d52aa38cae807070381489b63c6..8e5e0163ebcb43223df4ddf844ef7e5c32b39ef2 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_tr.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l1_w2_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:74a85cd35335791c54851d8ab3db02bdfacfae9e9a09b9d06861db8b7507c1e1
-size 2866563
+oid sha256:5eea24ec56dbee7822562670190fcccde0f9ec19fe4f7474ccd0b5f22267078c
+size 2868940
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.parquet
index ee9cab7384586513839e4c87eaf90f87e4fe4159..e026468736fc567062b954b2bfd956efac4d113a 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3a3f4f101c399c03f505fa6f2c17a8a40bbd5fd0dcd949df9d6e9962147a724f
-size 3583422
+oid sha256:2d93a90dad9343c3288aae674b5956277f5defae5951c3b0a4a07e79477dc2de
+size 3230144
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_tr.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_tr.parquet
index 6e0f85c6fbc7db3896a23cb0c7297fef5300ddc6..43921357cde2ccd5257e6e202c4d66407b040172 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_tr.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:51fa2799b9b04eabc7e7d48da284befe83dc5223e4d81c5b90e83cac1bcfb18b
-size 2866480
+oid sha256:a25947dde9dd35c917d0e764f33e2d068e9242cee8a8676b941c6496e3e97b2f
+size 2868857
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_natural_stories.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_natural_stories.parquet
index c9bae21c868646c9bbf23b42ff3a4546e785c566..5d3b04b70ecff4a87a42263ae6cbc69ed350b3af 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_natural_stories.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_natural_stories.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4c89c9ac5a8c634914cf72d6766e399ca845ce54e65248ece7263832b1e44051
-size 1280985
+oid sha256:34d39995f259475107511ab0da5e013b91a73c5143357b8a3284bb100a29fd18
+size 11824131
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_onestop.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_onestop.parquet
index 1d972424e5a78ed915069fc0c629016dec633e6a..196f708bcda238c05dda0720f05b7b929d3cd434 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_onestop.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6dc3522775990cd8bd0161a0c7752dd272b911dd69fa7c233ad898fada864d0a
-size 43120358
+oid sha256:34fcba90873c4ddbbecacda0d8cf931990ba5b2c8357a427f7824f0d2d262e33
+size 44393338
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_potec.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_potec.parquet
index f6128d1d7acd749a093a3ee09e662c7de1076677..927c762accb568293dc7f62f8a44e0d64b55e6b0 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_potec.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_potec.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7c40ee44b3cbfdc17e17cab25317e66305c19cdfa27fb08c4c7cc0f13021986f
-size 2978553
+oid sha256:a83c429c5c5b9f2bc481f0129e3def5035daf11d44bc034774ce2deac6894c44
+size 2975278
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_provo.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_provo.parquet
index 778432401eefcc52958171f71d12898c39e96fd2..5c2325acce5ef88874c9643556e4c10dd37ddc4e 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_provo.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a1cac53b9368a55dbf0ac64e4fe9867bfd2347ea4a908994968b6d6c14a9802b
-size 3983578
+oid sha256:760ba0a7190ecb4e6521b738867854fed8cd67114b74006a3981065dd0019f1f
+size 3959077
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_sbsat.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_sbsat.parquet
index d8e001a58c6acac866fdf3fb864ef8762441b525..91531e7492d5afd9b9f6a51aa62773a972b5a789 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_sbsat.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dd9f5cddfec624616d0090a6b76b8e436b43b954be9a7b70897bece4b5f6b425
-size 3590709
+oid sha256:869b8000a2d153ca792e3917bc1867678f049b9636d8bf2afb8e1b2fcc6696a5
+size 3582153
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..f0946e41d276176e09304e558542fb3303e9f26a
--- /dev/null
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a754531d9f7b51fcf29e33007f1063f61383ae3e8a8a5090c6fcdfc9a4087998
+size 3364876
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.parquet
index 2c30d3e3e24039e6393ef777b29aec72a41bbb32..8ec3979699f046c103810e246819a4ccb56e824f 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9d16d2e093d8ffdb6e3ab6f7fe25bdf85a97d54410b67e1b9c78a761249d946f
-size 4884016
+oid sha256:c375a235c71bf3b8c33d6c92267f8fdda00d32bb6f3c1fbfcb5fbf7294d1d518
+size 4881047
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_celer.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_celer.parquet
index f2517efe9a7414d1279b0d9febd8641b28699c57..18c778b36c1427ed242975cdf0fce3d23fd67fa7 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_celer.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:85ffb11003481b8cf8f428e9af2da9417eea1f8d28010af591c0f41c99103d98
-size 434430522
+oid sha256:dfd3c45b43135f5b532fca673c8dae89b15bf7fade9fe42bc75f7ee0adb96928
+size 433270489
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_copco.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_copco.parquet
index 9b94bc5631b77c6b62b116d366e2d5769a621a92..eddd05515a3a5fbf4414292e6cd361d8e64815a3 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_copco.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_copco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c78e4e32d22c18cba640fc34072e085e83878aeaf9774b519f8b33c350f9b98e
-size 40038842
+oid sha256:4ea0dcd5791c4ceaa7d207cc4831a0525fa8379d545349efdbf96e6b18c1bef7
+size 39963068
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_dundee.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_dundee.parquet
index a80d12b0a81555c50b7e874d4eeb950570dbdfa2..2888a1e5f5acbf38955b871b198fc9d9665cd4b5 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_dundee.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:706f40c4022164087935e099591e44a4d1df7d8a85c0d7b927170fdcff7143cf
-size 5578181
+oid sha256:e3ca74695cf5cc77b1babe5df94afbaa4452f48b459d2da222c250e9159341fe
+size 61417175
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.parquet
index f1efccc315fcf9cec456bc0381b195affb941378..2ecb3bde1264cd8d57b1116df823d4333499f289 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8ce1d0c1b110aa051fb48a5931c8a9f30db2e29c758dd15d2f9dc5db5b138140
-size 8509781
+oid sha256:4c2ffeb33561e8f02728c24b3677f21639a11153f7100d3c45c6b64ecae0c7e5
+size 8511087
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_geco.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_geco.parquet
index 966b98912d4fff5c49ab7264004b031f8ffa5efb..5f89e9e1f166cf16cfd396b8ffc2911e33173699 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_geco.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a416edec8691eff8d3327213f16d270842c14db680e315c9af0a1ab837de0455
-size 73165319
+oid sha256:3ed534c16c11f8344ff706a9f28868dc85189aaf1fefc776d0c73c440b92f99e
+size 73157269
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_de.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_de.parquet
index d1a0d5f169bf143f3a602115c8c8051d99b68ea5..4394e8fa00c3383128c7297d225e25d02d832baf 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_de.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:55538f18c608a27fafced5bfa05254547963f8f518a666846e8ede64523e3bae
-size 2952649
+oid sha256:ec48acb3e66a416e037ef9bb617a7827c0077c10894aeb78671c92225b7b7b35
+size 2944180
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ee.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ee.parquet
index 48076ca415a3ddb3f81ab0cff5ecdfdfd411cab7..f9c50460ffe503f6e992cc653d6ae1d9d4732161 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ee.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1473c1cae4880b55d8c279894909488768259e0f0a9c6160f976c0f5c8026286
-size 3306645
+oid sha256:94b6adbe5d0de8d69d724ea9d745cc1ed8a22ce942b5f3fb027e1129fd19e2cd
+size 3300098
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_fi.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_fi.parquet
index dc59ec497a74deb5a00df39ab2ef9d7d9074bddb..edc12c5ab290adb2542f3aab3d2390ae85004989 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_fi.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:12b85430e0598ef7546f5e99c57e53fba556e67f54a6e8f776fc118a583d946e
-size 2690067
+oid sha256:6bd715eec15515427c7f8458b93fe723b791ccc747586ef818dcd92c0f5c8861
+size 2678561
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_gr.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_gr.parquet
index 89507c4ab35a475fe3047f20a86d946e13165ad1..c255de5477c8bad6f99c706d2211c317172f7731 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_gr.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54a1aa002095ea690ca58bdc73e94d11330adc4ed4661df7bc793d1bd785462c
-size 3021905
+oid sha256:eaf2c65e7bf46d28c54fd9f87e7a3c1e0320456183d414dac0958861d8a6cad4
+size 3018774
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_it.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_it.parquet
index f84c1ff8972f7ab600d37aae4226a2a682063572..0acda1fc21758eac5bb9ed7f322698b0eb90f647 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_it.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d618fad657b418d02e29e95d2030bbe1054bc46736bb60c2f406b47ad04a1340
-size 2866591
+oid sha256:9f0b1f81709201a9b533f032714c374d44c2a00757f7db5be587ae6b7492fec4
+size 2874952
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ko.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ko.parquet
index 6d8aee8c20d625f8baf8d105c1127769b22f19ad..1b80b4c9013633408ea4cce5b16e64e479465fcb 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ko.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_ko.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:06bb18df84d519fe466ac4abcac40a36dbd94928270bcaa13d368240934c13b3
-size 3229398
+oid sha256:7e19b6cfb0d9514ff83b0adacce45f185c108778ceb746c2a9077590d13ab380
+size 3230317
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ba.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ba.parquet
index d9cf951d106f797e0f197f532e22b34632df540b..8fb4a4ec3b85b33221101a1b16b206a872f529c1 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ba.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ba.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6769b37409944c826396d8bf8da374f2d67a1ae86b2e3a9e6be1dfb3532e035a
-size 2972539
+oid sha256:d3d93878e750b6c1b659c52e0973fc18e4d49054bd95a699dc9ce1c0e4adf8d3
+size 2966408
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_bp.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_bp.parquet
index 910f5f7189a25671cfede49106ab30e372401a71..8429dc5c2d160d29d54c8064261125ef23719bad 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_bp.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_bp.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:380d1951e131b2620e14ec482e83a3110a336f4045ae2a333faa8deedd4ee93e
-size 2974387
+oid sha256:5c767ded50f06aa87d554d77b725f37d973a43035f2289e2bc6f50ddc5fde8b8
+size 2974834
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_po.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_po.parquet
index daeeb5370cc0c8ca322ffeda57f0d5cf2010ea4b..da9f9d835cb231b7ae03d168799e1f5463a985cf 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_po.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_po.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0cbd39131349e5da422a15b2061ff2c86060cb6ea8f1b994609062737bab681a
-size 2951869
+oid sha256:e72c63c5867e14c3d5cae097cd959c2e0b1f2b7bd470ac75bde841caec3e88d0
+size 2939191
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_zu.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_zu.parquet
index 135b0b478dc2914d65e5ba0e71be70cf7dd0a3a4..a25fd7269591089ccff61a81f5b20c77ae7df78c 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_zu.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ge_zu.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:31b9830452dbf90d63b914cb7dab60459b04bdd9d0ef34b9a1c4a9ef6a04bf05
-size 2951869
+oid sha256:9160d1ab831847b1260586c64220ffbc1ecbb9e9fdab971892067626fc58b472
+size 2939191
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iiith.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iiith.parquet
index b0f8cda30188bfbdefdb6e12b45d4faf4114626b..377d558252063e9be8959a721c4ac38159e77f36 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iiith.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iiith.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1746fcfd90cff00ec517e82cba670e653fd9f4f3e2df1db820ca9a6c5cd9ae6a
-size 453770
+oid sha256:3528e04a182f38afebf9a8d2675c62c28a10c92d80bdbd4a66c95f37f353ead6
+size 3033984
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iitk.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iitk.parquet
index c69a5eddb8b272352cc1bda089ee643820e0ebd3..c4b107a1afe980d0e9d1407bfa7e841ab3c8d2b4 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iitk.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_hi_iitk.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7b4a7735015e95660481f7fda77e6be5b7e78c7e2be20660c88b701c0b7b1eb8
-size 453755
+oid sha256:942236eab8cf99860b78cd93a6af8aa84fddb3d171412c200756818a94dd218b
+size 3033969
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ic.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ic.parquet
index 42b66ed7c389a37c531b8c0c500f0c557d6b8665..7b13ccd8be1d8711728a9c543afffc94f73fcbca 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ic.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_ic.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:846dd35260532c6c3dad1bc9f45ab493f65c68e7a3292db6bf0fc5ff3a27547d
-size 2205072
+oid sha256:23c8eaad5090871462cfbb353ee172053391f0fd2ad32568df9431e68e8944c2
+size 4300118
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_tr.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_tr.parquet
index b01611e1b6915542bd8ecd211b27849bfc3f1dac..76e57806bc96166f4ff65ff3f659fc499287ae66 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_tr.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l1_w2_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6618bcb97501fed17a67ef06a9e69feb8376fd8ddd2243a39e590eb882713b84
-size 2866552
+oid sha256:eb85019cc7f55724ca83abcf700537a716352a641458e50e18b33025b4ddd356
+size 2868929
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.parquet
index 176709142a3115c39241b0efe63fcc6af3f9404e..fbbce3140dcf8e7a37ec50ff571b09733cfd0db7 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aac617761bedc7f81401815cf2510b7c05cac8b705dc2eb07061988b68e333ad
-size 3583412
+oid sha256:250a7d04fbfafaf586ac3772fe668ae3a4a9991f22531fa8b5f97cdf177a8ee8
+size 3230134
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_tr.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_tr.parquet
index 360e638794b067cfb6c82b109430303498027f46..be7baabfa11aec26e85c8a76271d2bc181391bba 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_tr.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8bc1941768bfc08b83638654a0ffb33459c540f06e1765858e4a28918bbeb3f7
-size 2866470
+oid sha256:a93517ac5299d4aa8e66a31c39b8cfc7585ec4f4145d474673fdd8fd8a432d20
+size 2868847
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_natural_stories.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_natural_stories.parquet
index 9a2c7ec8c3ce1e933c0f9afbe9e902095598a1b5..69810f7a502ff064bd8029c5e1c9e6354c73c617 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_natural_stories.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_natural_stories.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:61660b79442942d02402e70e12502d96832ea010a4d8f28972858bcfe9e40eec
-size 1280975
+oid sha256:8031eb53ed905dd70aacf00a40220d03833b2d2741eb30c147733736c0882ee6
+size 11824121
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_onestop.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_onestop.parquet
index bc8a78f2ecb2ab0f29fc7e55f2546047f0b53feb..66ebadea242a17286fed8a186e002b09cbdc9561 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_onestop.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f6c1d4040160cfa8fc80225ef7820349987ab91841024e917bb5d43304bb7337
-size 43120344
+oid sha256:e1208b8dfefc610e5d8feebfc3f13676c48b7c16a8987760e8d0187514551c97
+size 44393324
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_potec.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_potec.parquet
index 80e96c1e2d84ff9c34ae52505c77a00c91fda29d..f6fc747d8a051abbc549c623e6cb99451aaf7cc3 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_potec.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_potec.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cd0eb1704009e717fd6b6fe5d6b1e640e23872abafc38ff28a683cd4d97f8871
-size 2978543
+oid sha256:ad00bf1f3e1fe9a00434e4d6e3a08d54bb63485dffad0b9460805d867b4d3176
+size 2975268
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_provo.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_provo.parquet
index 854636731427235fdd99f87141d04be79c6ceb7f..d87437ef52b6b1baee46b8da5340d11400f76073 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_provo.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b029c88723e2cf98c006a48f82124c7ecea8af10237b63fe680a711c531776af
-size 3983568
+oid sha256:9412abdd18fde0247f5e3f203aba5a0514c391540c2bee34ca1d89b0c7f68a76
+size 3961562
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_sbsat.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_sbsat.parquet
index a15b3a5df7c815cf639dbbba996345017d5c243f..cb8013b8e8f658bf5e4411f304eb9e5672298e37 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_sbsat.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:820a8a905340706d2a583bb23e0c0488f0668507b39ad99891aa107780e8089d
-size 3590699
+oid sha256:37a18b7256af4da415ab288ac50e5bfe7417df6d0a1a6ed06160f95c55fe4b0b
+size 3580082
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_story_sbsat.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..3a045711b85f84d07c00660ccff9d26386c25145
--- /dev/null
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e632ef6ae0fca04c4c22f3ccbaefa663ebd253a10c51725e32965d5c895d698
+size 3364866
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.parquet
index cf91ff136540d1b55abee445f913005b05429071..638c6255f3a65e0e7e291201762188d10ab19151 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b844382066b25d01549b1698e90df9f5247291c9a835d6b5346525b2d7f57fa5
-size 4884002
+oid sha256:f2bc638b4e0d6c3724bc39927cafee2e8af304943dca42bd0bf82e4078efc2ff
+size 4877372
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_celer.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_celer.parquet
index 501a671a3823e21f3fcbd2c613505f4cc1001afc..59596067b501a07618f1fe89108729ffbfb09e7c 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_celer.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:00de5414717eae533ff5820841136222a5e47796e50fa6cbc9b8509b296c02f6
-size 60945207
+oid sha256:840c33978805a9f50a9378840050b6416e7f9a3c897ab75eb0cd051ddcc69eb4
+size 60898287
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_dundee.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_dundee.parquet
index c57e662b96c9a3e4bdd86e30434879a847605995..c25319e1ae48d059a6e76eafd839e6f4f96efe84 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_dundee.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b6c3b413eb2446ee653ed1c81f2626c9b2178788857cfe1d9183d15414fdc77e
-size 5467967
+oid sha256:7def4986bea3245b514e5c131ea3824da13a9d14bef20a717587c0098249bf26
+size 9569374
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_geco.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_geco.parquet
index f796aaa0849bbdb8a27361e36e1318baaf36506b..31fdea19105b1d617f1ba13d0759890b7d4e1341 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_geco.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3b80bc7b634f2a166a4702a413be90832bc9251ea4b17d7bb374925ca9bde998
-size 11064769
+oid sha256:1a02fb29aa916a951d50abeb10dd84a194011abc2662cd99b2595daf41ed8dc9
+size 11072577
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_de.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_de.parquet
index 8dde31cecd5076297e9d6e39d5ff832d54a2d13e..70ed8ada88c22103b842b8fb490ebba8e5f3d951 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_de.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2ea73d7ba065ba426aaa81059c881e610303e3ae00a6526cc3bd4b10a969fefd
-size 447807
+oid sha256:a7d1f2ff40d0defb769fdeef9bef84d7b59e4820db2b302a1b780e2d2b7bcc76
+size 446808
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ee.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ee.parquet
index 4109dbbbde353ee13951ad2d3ba3471452cdfa7f..24fb28c38fb3563ab4786f7cf95d10542c4ecded 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ee.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e990f9bba210da08d47befa4559993f5dc865be422c9839ac3308dd72aaa6177
-size 531093
+oid sha256:ee5b5ae301992247fb4f0c25f9eed0032ab2ee061c50c42474d91f34e91162c7
+size 530208
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_fi.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_fi.parquet
index 559c448f3a7e58887328f8e6a5eea893fbbbbf75..19d85a4ab0d13f010cc6c66aaf2c9e605e993a2b 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_fi.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0db7eb176de9c30d5487cfa16dec5beb14a214d5fff113436d07c1fcc9f3e3e2
-size 405119
+oid sha256:677a4d2d1520086897e65edefe117933639f3bf8b1bc60a2883c1980288b473c
+size 402981
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_gr.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_gr.parquet
index 8549f066be89725306bcbd20f7807361f533ec2d..679066897737d0e3d0e3f83d11d9fc0b7e104613 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_gr.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4b5dfd958f547ce84d9443b37d0caa2e2333f44ac55980d2790fd8385a832a0e
-size 459766
+oid sha256:1ffbd861422143a2529955fc22a299f319df4d2075e576fae085c547bd3b6e76
+size 460960
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_it.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_it.parquet
index 675de44ba66faa318ba6e43578aba6786d98fd18..96454747536ff5d596650c392e8ac0c0958146a8 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_it.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bee95e54cf7ce91d0641b53fdf0765cf0020c93d217a948241e29e4f6bb0343a
-size 433086
+oid sha256:9c57fa31f3f3c00fcb5986b5812b9eab7cc4e30c5c440c34ab5fea44f0c818d2
+size 437284
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ko.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ko.parquet
index 606eea5d1771f36ee9bf27e442f8ab2bfd46644d..092e4abbec5ae7e409c0645cd8dac850ccd6f9d6 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ko.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_ko.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:273d457ad0255ffc5c4db97965ff8985fec01ccde2b1b3421b17b25fc7981bd4
-size 533180
+oid sha256:5fdd72bb1d8d4fb4964cb29647421d9455e477c989b92eefa1ead41076431b8f
+size 535308
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_tr.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_tr.parquet
index a6b9d2fa5f0fd966ef44c5f93dd3ef98b241f404..b3ddb244fb2edb5c37663597ee2fe334df337e34 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_tr.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:26da5603bce25956867685ada2630d625a9830cbc1db03431e2be7678ce396b0
-size 438944
+oid sha256:7600a10179fb159d50fd544bf404678cc03fd372da94d33a8936ecd274a9e097
+size 437918
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_natural_stories.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_natural_stories.parquet
index 8fa72b5ccdfda8bd2333760df95b2abfa4a43bc4..da8f15ec93f613a7625884c1914ffdeffc6fa3fd 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_natural_stories.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_natural_stories.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54e0ea96fb5af2b8cdd4c42da28cd151bb9b176a10c8ee86a9f48b0bfd88c85a
-size 1308901
+oid sha256:ee8926559fea0ee26f4176ccec46b399dcabd6f41243f879e9b38be54410de9e
+size 2282691
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_potec.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_potec.parquet
index e2718608d2c0a72b459d7b4296fc292f17f326c4..e4cef7045f790dcd2050da02c84cdbaee638f809 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_potec.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_potec.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:680242f819d812b565f2eca27c55ed7ecd1e2354e928ea1b6b3717315052c4ea
-size 455665
+oid sha256:3500289c37f8ecaf7426ef7310f72aec351fb0cba9ec7df568594654926e9bce
+size 453623
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_provo.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_provo.parquet
index 56c8b3dd542d97b97128221c06a240b4261057fc..3794bb51ec314bfeef5641f042b69c74f89a1360 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_provo.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:174494d87e86abd29b35dba1190fec5a18e3879040e6cb73d6fb148bf607ee49
-size 686799
+oid sha256:55e367af54ecced4b802dd42c65b34e8b7f5244115cc7a512c77d645d30991d4
+size 683235
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_sbsat.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_sbsat.parquet
index 1817f94f45d50abfe9b4eada1f2efc1d6f127b69..11070d9af1817403d7c747f78362c060101c3bd4 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_sbsat.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c4c55f6f5eb143e4ee6136f24f837915cb679dafaa57282fc408ccf2a921ab3e
-size 618836
+oid sha256:56291ee135ad22e6d28b117b4d7f7e1d6e6cf60681bb29e6a1a944b7f940b648
+size 612151
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..e8162adefef0038bae30f1e19fb2b91fc2354bba
--- /dev/null
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1aeb5a15c50d35c68462d7044b6620d6e04c9b72fc3010e7525a7674bc3be199
+size 578394
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_celer.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_celer.parquet
index 71f99268c7ff7d0b01ec72750059daaeee1c88be..02b3cbbd7c8162cfed4e9975080509cc10cea9f1 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_celer.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:06df79bc080a186d23c56013c04cf6405d68a5f7ac1c70077987d76eff417930
-size 60945132
+oid sha256:48616c2daf78d8acbde75d6c4de2cdb23531cdc0a2789655bed9a721bbff4003
+size 60898212
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_dundee.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_dundee.parquet
index 8fd763b990bbfffa43a2a9ac788ee15ef9f5f811..8b19f0dd6a6d7cb05ac3dc7ece125f9980cc60c2 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_dundee.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7b0aff1f2333b45405db00bb321d26c26051760ff26eaab7219bf35180c83354
-size 5467949
+oid sha256:92b402b2822a600b5124ce4aa18584518942aa1ebc8ae967039b4010a2e75075
+size 9569356
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_geco.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_geco.parquet
index 77af588bb568f5d40c784b00e6a7ff62d4187644..183f2854499209f683e61f44bd75605429799f33 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_geco.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6e4df88103b3476553e10e7aebf438c97c5e6e6be22c1c68c3dac49af318f4c6
-size 11064750
+oid sha256:d42b6a8b37d43a6ad99ccabd3c54c82988487955a4ecdc2b93f59f1a216a9159
+size 11072558
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_de.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_de.parquet
index 91f842d26e02e97f4c213676e19e76919ef30fba..6dd3bb00d0b4243cb9f5fa2e6a6d4c3bddf22744 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_de.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:16294a23ae6620c8b15ed939f02993115724874ca06378445cb365c8ab9bb2fa
-size 447797
+oid sha256:d4840023e5af62f5260b2c0c9e11090bd5571aa070b1bd94dfa705f756bd965f
+size 446798
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ee.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ee.parquet
index 383931fdbf50402b2cf8c86862b54861fa959332..e0cb08cc65529d6ad35744941e3658871a3f267f 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ee.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4138cf0a8665095d509e4832ef3fd4bb62e0671bf43af77ea2f89a5eebd2c0a9
-size 531083
+oid sha256:6c2238ce9f80b7cd0968667660fbe47ff9c5c741a50dcd8b256284ec92f152b5
+size 530198
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_fi.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_fi.parquet
index 5b384a08ca52696d5081a0dc0f02d38017b68213..9fbfcd372e5b2beb4769fd8a09f45b27bf1cbccd 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_fi.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3957a76cecd316aa864b2739c08f000bb90d6ee2303627d9d9b8b6efef46c1a5
-size 405109
+oid sha256:c3d7198e3dbed3b19f6ac15106d0b761f40319199e28b3839caf22e8e8bcb839
+size 402971
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_gr.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_gr.parquet
index 56f2a06c9a9298d781e240447f3b85f3ad0f58fe..e1d7ff68bef37f7ecb115902d11ffa9e3cefa95e 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_gr.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b34bb74f7cb2ea2418222303c1d8261683ed71428686ca915e2c418d876018e3
-size 459756
+oid sha256:3ca416bde44d14b3cce0ba00880ff1e9a60d01b4268bf87cb08c996a70deeced
+size 461124
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_it.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_it.parquet
index c82b1ba8971dd260359365a7b1f6c7a182d393f2..348c4a13ed6348e6435a1bec3a4107437233b26b 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_it.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:94a3af148e76d825a99bbe9ae9ffd1d6645d6b4b3b26efde58a252b0b46320c0
-size 433076
+oid sha256:1a74e12c6ec34514304386df96e1292fff209e2cfb7e73a68b2df35b01ad83f4
+size 437129
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ko.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ko.parquet
index 414d54e79e5d6263a28e89a524499ca84fbbb63d..0ffabf3c422b15837afa3d927e19b63f34e3173f 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ko.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_ko.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:39b140766660f6df9040a03cb0967a159093c6f8ce737d6a730aa97c613ca245
-size 533170
+oid sha256:7d57aaa953f79770549d26cd8e7e598c8b2336dbff73c092c4733fb0bf4f2db4
+size 533914
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_tr.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_tr.parquet
index 2bc83cad78900f6c70e7346e384e39e6aeeaa824..a7845579c9501d0eee60c367c4dadb86867f45f4 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_tr.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ef310f1d0f7731577a3b9f1d011ba1a056f54c9c9fdfe0af7d25353c11cfec4e
-size 438934
+oid sha256:f56a7678bc121f09d874c51f6c73fa0ea9291281af738c9f23fe2fbb6062f48d
+size 437908
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_natural_stories.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_natural_stories.parquet
index 106d589465de8251286f482838180950a202deee..1efd6738b448c846fd5059f24a278dc2e1d5d80b 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_natural_stories.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_natural_stories.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0980a23146cde6d0c9eb0d16dad3be30f999fb9c35fddbbe238ba9bdce488f0e
-size 1308891
+oid sha256:df4275201b2d742a2be06abdf0e543de6ba402986812be4d3d62ac4bcf942954
+size 2282681
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_potec.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_potec.parquet
index c0a6882dca4007a2f91233655c6cc6db219491e6..2c0cf5f5b236178e97d5d4aa97f5aab6a869181a 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_potec.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_potec.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dec91a1b633e829b49473c1c9781a2c3e33b07479def27bbe59a3e9a7afd14a8
-size 455654
+oid sha256:d436670f74ce2cb2c0d4863b35f25ada7fd6b57a85653c0541528f5074be6527
+size 453567
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_provo.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_provo.parquet
index 5d8895a65976a263320d835ce108fa98ac2b2bdf..1acdb667905cce6051bb458289fc15f62bc21eb1 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_provo.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9033719c1f3242cae49dd6bf22a58335ad1218e9cc5c1ef461e7925b531edf8a
-size 686788
+oid sha256:d36d74b1e7cf5675de9408de29c7eda1286bd3b7c3995e37900894c6ad75d824
+size 682640
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_sbsat.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_sbsat.parquet
index 30185a29a40fdec116f9e5dbd9dbc25e203b58b5..6d59e119011553a4b789a2655fe7a468d2ede808 100644
--- a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_sbsat.parquet
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e7007fa763fdd0db25caa0503772fcefbb8ba4a2ea2b664c1805ed51ae783279
-size 618825
+oid sha256:cbd171a4d1eb3d7fdd426eab534a0400dd1893d16812442dd1075608ae897637
+size 612140
diff --git a/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_story_sbsat.parquet b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..0ec1a98d2384a9b9bb600e30c5a9255ceab9c1d6
--- /dev/null
+++ b/mcword_samples/ai-forever_mGPT_mcword_entropy_N64_M20_a1_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68b6e4291545b2f4eb3c540b8fea75fb5bce91734f9b172eafb4523a9b439297
+size 578384
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_celer.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_celer.parquet
index a4d867151b91b4d1dfd9c50e586739933dd93eb8..43ca6c5ae8075e81c276764184c20ed2971fdd24 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_celer.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1210d6009713e01e7fa0fe422e9bb47ef4674a646c7a9357d4ace78827c2d186
-size 449890944
+oid sha256:c6b346991e3719059cc94cd2c2e67c6b18983464bf627ad2a0224be4cc528a9b
+size 449220423
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_dundee.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_dundee.parquet
index 5ab52f4da7d45037f0f142d772a420ccc0d331d5..6e40497ad294d4b33ecea31ca1a37f7885214910 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_dundee.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2ab7643ded4ba574707e0320f1e081fe48edd0ecc073a15fafeea9cd8975b110
-size 10063928
+oid sha256:70dd1bb4986a784ebb35285c3cbfbb64114b2b7fcfef908e60ba9326768e539e
+size 65252064
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.parquet
index 2e279c33c7a33042e281c4bdc609ce147a09f48d..c6b4c956ea17e343988eaf35916fd911ba516254 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_federmeier2007_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ac774cdb8ac035f2dc2c3287cbd4f62520ec0b7d4762116c9bd8c949b0e53d32
-size 8890506
+oid sha256:d38edf630a338ee5119d7ee3a107520714db2487ac3be3103bde36083308d77b
+size 8889974
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_geco.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_geco.parquet
index 086127ef299f83f19ed139ce0a9b999f17e38583..716408575b7e082573c0de27054c3dfef195ff03 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_geco.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:578d02d94e42e896d60ea733059973d7e3ea5ba8cd63c310e934f9ea7b79c4c9
-size 74498680
+oid sha256:9be8e6c2ba05b72d78db3506c846e11e9401717b4f990b383e3235d5dd283161
+size 74464084
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.parquet
index d57783b5c50b40b7167255b0f80a0ed93dadc1c9..167a1c60677a1cbd971cbd20e64c0c115757a071 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_meco_l2_w2.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a1cd180be15114a4cc12903e3628462dce9750793e970f28521c9f4858ad6c88
-size 3827332
+oid sha256:bad453b452d821c27ff67f9f8bc2ab6e3ff176ab5eb1f1c13b2224290dfbe4ae
+size 3361904
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_natural_stories_maze.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_natural_stories_maze.parquet
index 465cedb638e24338b3d6664b0288caff18ff8610..f33a060daa5127247cd3df853d9c01ff5a997ee9 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_natural_stories_maze.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_natural_stories_maze.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f3b09fdf558791585841604a5050044bb997a99d5ad6555a02426c03e538777c
-size 14385409
+oid sha256:ed216fac08d28d5c9f0e03e01cd35af27bffccc13ddd6700c1cac5d8d6d600d6
+size 14382955
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_onestop.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_onestop.parquet
index ef6062f3c37512901ff169ecb7a7d332c3bcbde2..8fa4a1e9da365d45fac9cef91455f6518ef0a361 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_onestop.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b704867246f61415800487b1beeba3e61512256d60b1936e12ead9232593abdc
-size 44420044
+oid sha256:1ccb7e54b464b3566f32cc44bac85b2b2c8b1e9b836c7ee35550327c5a7641ee
+size 45753739
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_provo.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_provo.parquet
index 00fbc77c2235e32a4837b694a71a17d47d9ecb47..34d28e8997f4d2f0d4932745bd0bc159718a52d4 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_provo.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a3247457fde064be4274cab2f58c7a739fd24fcb267a615d3b9d60e2e5907602
-size 3991580
+oid sha256:9dac257f7f94220e47c21cd18bb45a40f2f68e5736f8c8c402ce46d8cfff8a26
+size 3974035
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_sbsat.parquet
index 7e66320199c9ab928bffa758857bad99bbab9e11..02f0343b65d530bdc4656eec85ae925a504a4d7e 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_sbsat.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:014b7c75c2d808f34e908e1b6a514f7394a3e01dc74d1af03992f7e1ee041660
-size 3692462
+oid sha256:c4bc62ed8fa2a567ad13cc404fe5bbfa4515c041d5baf68dd3685a20bc6b8972
+size 3679102
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..58930429ae78122fcc339c3c20b51bfc54585720
--- /dev/null
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97fbc18b7f6233b5b9324a8ff7a359f5c2f00069e80eb214a39575d06a79ca9b
+size 3452322
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_federmeier2022_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_federmeier2022_n400.parquet
index a83f4dd98bc2644e2d757867413e1981d34ba30f..a109cb55b4f7a12dcc87febf8aee59ace028b1c4 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_federmeier2022_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_federmeier2022_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a358b3123f9844bb639da813df918f1b9d3216048cc8dc16555178d49cf133c4
-size 9363818
+oid sha256:e23d58646334796fbc3d8b98f68fa92a293b9800315593883f8618961c41f1ea
+size 9374644
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_n400.parquet
index 8e7d32d84323d9ab41ec52885b6141a5cb4ca62a..ab446de15996fa8aec3c584791ff0717d2cec5ec 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_szewczyk_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:42f3afec56eba7d06d75d3fa077b55b2c210e91a0cea1a8685d6fa928838a62b
-size 20657941
+oid sha256:6fc32c09239341dad82fc89ae3119d29575fc557b8f05f0c6f972647d7ff5a5e
+size 20692890
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.parquet
index 9cbb05dfc234eb5758fe692754c8113d9c9299fa..fcd0087bbfdd1663fad8e7975565ca20f6ca97ef 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wlotko_federmeier2012_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0607893c736326a521a0fd0f992394ffb72d42556ebf286151ff0d843d4e759c
-size 5078580
+oid sha256:3a3d6be2ddd6ef28b10c58ddccef90c0d105ca37996a698bb29e381e78a989e7
+size 5083291
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_celer.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_celer.parquet
index 454ad2519ab0bb7ed3c5370e246b9401e533d166..e3b0d671590087e841217d017f00f2e91a9a73ee 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_celer.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f56979ac2712e0536b9a4fcca1337ab35a9426a7f45caa117d98596278d589b
-size 453915066
+oid sha256:658c302e59b5a8bd159ef7e61752bbb4088be8a8ab4f5b2170668cdea39451d6
+size 453052945
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_dundee.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_dundee.parquet
index 2998d380d6ae05fafdea601c88ff32df12132aac..d94c5d4ca383f7d9a013a3606e9aef06f20db988 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_dundee.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:213d5a9c87be39edacf44c4dad86c3f998ab68f7cb9dacdd7d39ea7bed16d7ab
-size 18450539
+oid sha256:97b37d14fcdcf1d2e7a0cd43dac351405d4cf19747a21b4476eb0870764beefa
+size 65358340
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_federmeier2007_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_federmeier2007_n400.parquet
index e8dc47493a8dcf77ecc10a64487605584897bd1d..18c0355555afeba41d52a34af6379c8b42bf2992 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_federmeier2007_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_federmeier2007_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6600b4aa6e1dee5f82cd728c22085091bc8ae2fe9dc75ccabf26fba77db9c446
-size 8923430
+oid sha256:db3e1123acc5f0aebda06879b1f414e2d306adba81ceacf3876da3542f0a2aa0
+size 8936118
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_geco.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_geco.parquet
index d4c98be5d122836f26d468f6814c44e1228c2963..e9aadf4a9e79d3b02cd170f5d7bf090e6639452e 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_geco.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2a2df1a7cbedbba0f8b4501525bf349b78975dd987fc8a24b12d203dea30ac9f
-size 75189536
+oid sha256:7ab181d37f6ecda6f18b3fd9e10f34bc0ee5f948ed66cbd6e65acdcb66b1d0fc
+size 75157437
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_meco_l2_w2.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_meco_l2_w2.parquet
index 6c527c24e817b4305005f591da0313b23860cb8a..66cba57e1efe6cefbeb99d5daf9e396e8d524db8 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_meco_l2_w2.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_meco_l2_w2.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6d3c62be1908cc52dc0ee6aa82407ccb3cd7f373dc01596900549dc4532610df
-size 3873890
+oid sha256:51fec1aca8f257618f7373f3119839019e9bb1591311eb886ccd9f670fd718df
+size 3393384
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_natural_stories_maze.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_natural_stories_maze.parquet
index b129c97599d86a3d8d85d4cee11ec2676d294b5e..f600fddd133a461fbe4ef326296c525c29768330 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_natural_stories_maze.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_natural_stories_maze.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fcdb39165689b84fd0388fcf8c6db13430a11555894b439ed786de7475f5d740
-size 14509760
+oid sha256:24179a921908a79d829ba164cf6d1c11c103368a751d2b3b7eba72b435d0eee8
+size 14516401
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_onestop.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_onestop.parquet
index a36b0cd90071db1871cb6712478fcfc1f6f904d7..ae7e14e7c6ee3da7193eb27106f93ecb22b61155 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_onestop.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c5cda202e7098c853d68108b690c5acf91ed87d025203f987c5a56711291c0ba
-size 45476151
+oid sha256:8b5fdbd84e4b06515177811500b833b80861eed9b9ce8cb6c940c57b7a469033
+size 46815088
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_provo.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_provo.parquet
index 84755ea771f42ab8ffc424a9298ab63946200add..b103346acdcdcd8bda58f0a67ff30eaf2168f0e1 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_provo.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c28e8171a603a6164cf7180e9af29a5ae50207b3b8b553793289a1bd9acda93f
-size 4047659
+oid sha256:f81a8ae79f240577208079009172e909ab2a229043a3218d5fe365f99b0871e8
+size 4021099
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_sbsat.parquet
index c107a51bffa4e573b65bf19f96ad6f9e8e3b21fb..9a2556a36d56eb39c8dd2cdab2aeb85d8a29638b 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_sbsat.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c4775424a5627786451d7ae26bfb45aab9a1fbf9eb09b4e7b12d29155cd5ad16
-size 3729890
+oid sha256:153d88f38d0ef50da8d3342475d43729fedb75db8fd8d2a5c2b9e95d8d7134cc
+size 3709734
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_story_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..2fee75d5c9876693e8e59ccc090a2d9644a19f56
--- /dev/null
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c9d0bd43af8b43d1ee5258d33732cd5668548d7d7d875c2eaed7b1f7e12ce29
+size 3464992
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_federmeier2022_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_federmeier2022_n400.parquet
index 3c0006d83b839a14cd8a8796ba2c99ac9d64f78d..f3ef97501abb06e17dfb9b3b6ac4663929192df2 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_federmeier2022_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_federmeier2022_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9341cf498f87daa896b3a2bbc014ea18eb5ee5d32f8f2f242f5a7e4d188bfefa
-size 9405308
+oid sha256:fbdbefe7768bca82993a4aac914a8c8d97509379d9d5d0b6a48c718c5c620040
+size 9413753
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_n400.parquet
index f7d6503110db932edc5ebe390eac4e586f85ca62..2d48eb13c4233354f91960621a2b7b05045475ac 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_szewczyk_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5c24e314daca4e66271b96665bec7ce9fd221cddfe5802a877491575d526a036
-size 20785988
+oid sha256:91ca6050ecb26a0fe2aa578a80252b755b951e927c13eaa841e3687c1039e196
+size 20799723
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_wlotko_federmeier2012_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_wlotko_federmeier2012_n400.parquet
index 060684d21aa53d4628b096d95a6df81de82d1743..4e965c4ea56991e2952e447ef0d55d7b803fe0e0 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_wlotko_federmeier2012_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_wlotko_federmeier2012_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8bad5887828df92ee439210728eba30b00e0f42c9043f8bd045e520fe2fd2004
-size 5101131
+oid sha256:7ec690c6b6eff7c23c5363f627ec7951fb0dc63a2fdb52677b8b75df2fbf75d5
+size 5106042
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_zuco1.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_zuco1.parquet
index 2bda7da9c7c66f11db46724a3ba3dee8dbeea582..cd13429e45579ea1bf4b3ea92ae1a39f29195d8b 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_zuco1.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_wsg_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1463948841498a3c4b6635d0d23da5be26214b15f0293277837fb324cf8026c0
-size 9133434
+oid sha256:0487b9c6e63a6539bd03bf46123ea28dc0e47acbc7097ad59aa976f8711dc7d4
+size 9150471
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_zuco1.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_zuco1.parquet
index b6272916aeeed609131ff18d53b09262471645ab..3ebe8f449e9280f98ed804a7c3fda2b725ca9aba 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_zuco1.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a0.5_s0_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5d68dd45be4654bfcb65c3f04e526e4a6a79199e2b69db2035d34c4368203648
-size 9071717
+oid sha256:3ff11e584e34a648d18ac7ff59885eaf31121cffa1058313a8de9a4ee7a304b4
+size 9077190
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_celer.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_celer.parquet
index 51d023e5c915125cf799d05a7998dc873e6a3fda..058965b23a7ac962ac2cf9156ce48f5e87a2f9ab 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_celer.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:25a6700b66fa77365265fc5bc381c973ccf2529e27da535cec431b4b49c30147
-size 449890864
+oid sha256:b7644d87711d55acda37b7bea50d710fdfea3749a9e932187a40138ef53aff5a
+size 449220349
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_dundee.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_dundee.parquet
index c804d5c8d95619cbf09ae5b0d17b5cbb35f9a917..abe8ba2cc80e8bdfe2c4e575a3015b42ecd6a8e1 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_dundee.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5e6f8c687524b6d6f6f73a34ec3f5344e3bef782fde821f8e6d8c09ceb96d395
-size 10063910
+oid sha256:6f4586377e6085e2528bd8622451283eaa85e6ff20722a45ac3a8c405dd3b36f
+size 65252046
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.parquet
index e88895711a260134e0fdde6e9a906d68f32b6a19..4c8489f9d23c91117df967ab101b2b7beab7c9f4 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_federmeier2007_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e5e701ad59ebf11a0502d914cb9c234d22397ad8e10532bcd941de180068bbef
-size 8890496
+oid sha256:04b2e82285fbeacf5ae126f406751bd46d1a47cdd29ecdef10d1ce933e761631
+size 8889964
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_geco.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_geco.parquet
index ba04823ad2dfe4d507cd96a6e3a0fe0ed6ff3e84..cc4855c84c648c137759b5753bfe8c8d41e23b8c 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_geco.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:395cef2a7fe098afb05c037b3b8ad10e50eeb008e213322a53266c03dec2ffea
-size 74498662
+oid sha256:4671e21bf480642f75a000c5c39c617ebe0c6381d730e9a26eab981967f6f9e5
+size 74464066
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.parquet
index c7b431d5bc9eed3d2fab013fed6702e0f3213496..ded5c6c7b294d623dd8874c66bc88bdde1ea5787 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_meco_l2_w2.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:62dbd1a8480a036ff86346b43d378b64a82f8476d58e7c6088796d43c2e129ae
-size 3827322
+oid sha256:76a71cd5e14a2dd99f7b428507caecbfd633bc865c627fbef7bee5cf7176fac0
+size 3361894
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_natural_stories_maze.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_natural_stories_maze.parquet
index ac5b569c53039e307a4d8a682de403ffe37e6c0c..3cdc08ff4e4c6d0854c4cec87ec21c8158559b53 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_natural_stories_maze.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_natural_stories_maze.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1958cd1a28e89c470721240c00c6ceaaf763ef3edeace67981727035276c006f
-size 14385399
+oid sha256:51052b8bd42e8775ef96076f8f51ac40672b22c8bc4e844d6d35182bd3e40828
+size 14382945
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_onestop.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_onestop.parquet
index 1c852c00f03366b5ee4cb04a47c0e77cf560e4b2..72dd9c9c4d1e6044395b9c6f0e6ad45e82e0ea76 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_onestop.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:427d6c5c7b551062099dcf210db77fcce6ca2ec1171187b594ee997f885cb1c4
-size 44420030
+oid sha256:efa8487bbee50446c3ad1691930df224655d22bce459e44f0c3fd152702d404e
+size 45753725
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_provo.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_provo.parquet
index 027578a4403971a01e50530e939eaefd3ba28a03..e0f7b64eedef9b630f7465b422146180f1caadfe 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_provo.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b444446558bbc81730c2153ff57bf1529334f15dba9405b4d35d57fb6b84c7a2
-size 3991570
+oid sha256:b4eacedfd7136d37fd3dca145a95bab9b83c35375fc58baf93d3a6a31499dd56
+size 3974025
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_sbsat.parquet
index eb1bd6542893cfa610bb2167aa30f446d3e58b46..16507580e3ebdd2dccfb7b7c1161d8715209f7f8 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_sbsat.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:46480b690b53ae4f40d9fbb6de95a58a03dda1d5bc8b8307845873a4d8a2fff2
-size 3692452
+oid sha256:dedbf4d113a588c325bac819c54ad7246fe138cf5327f252cc3c15e92cf4c60f
+size 3679033
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_story_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..67ca4c491283f5a7a7740a483d06ac1e6432fc25
--- /dev/null
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d66d90d17618fabc45efa7ee62ff86f3cb1c366da5dc9a6e6e4f8c1a25267ebe
+size 3456462
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_federmeier2022_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_federmeier2022_n400.parquet
index f921c63f4a5299f121f4c220446607c521f77cc0..5c2d2fcf335c813134944f7ec7d608064c1bba05 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_federmeier2022_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_federmeier2022_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aace91e2a3ba78a0547d7d51ebf8107467a478fc9df93d58e8ca3345ab6ea5af
-size 9363808
+oid sha256:5c403f9da3cdebc160d6b6b0e884e06457ad434915970b3742d8eeb48bbba3e7
+size 9374634
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_n400.parquet
index cc4da25a8ca620528ba8b51a905d74eeb681dbba..0d84e4b32f4c0790b3d842eca3c0e29642acd31b 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_szewczyk_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:34d46fe746c91188f20b131a6d50c3e24051687ed88fbaa21461c1c13386ead0
-size 20657931
+oid sha256:ff70af92a86c892f742110bb597420d180828a56ffcf925a20e98d06e0f2d2be
+size 20692880
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.parquet
index 3e2414fc21897b78d2a6a2ba8518cf02d4c2725f..ce253c85f8e68d91ea44be2482d70357bb92d562 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wlotko_federmeier2012_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:192ff6f4234bb1aa6c7be2ee490ea8a864e995d8106b257446d0a84def689412
-size 5078570
+oid sha256:b0d93df7c09cf85935531d01471e2876a192b02182867ed8faacb6fcc46d4e94
+size 5083281
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_celer.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_celer.parquet
index e0a65f7c9b97ca5f306dd69ef11dcab5670d3d1b..a4d90a2d47550e8c236171fd9b3f1e5d43058d2c 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_celer.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ebae2526167bfc120d281129ff98fa30ca516639d1ec601680d18571688dceb1
-size 453914986
+oid sha256:4e11a4b5d9bafee33eeeb1678e588892b75f81613a73bb71884a036ad5e56420
+size 453052871
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_dundee.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_dundee.parquet
index 7489661801d462a878c51425a364ea21e1d9e15c..8c70223535eda584a51920f66c0883a2a7a6775f 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_dundee.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:951d6862070d44e513b3b6875f81e8cfb7f65c56e5b0cfecc087bb9f2050fc6a
-size 18450521
+oid sha256:5f470e718ad250145ec6de0eb6bc57056c957b7248b13568a7515ef5b842a187
+size 65358322
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_federmeier2007_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_federmeier2007_n400.parquet
index 34c545fd1172ce1e76138e3438ea22c1f1dd0bfc..ca7f774f7fabfeb3e671afecaf72354fe2a59f8b 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_federmeier2007_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_federmeier2007_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:003b92fde3bfbdbf17eef3a4497ced66a082f7e18a7d40e153680a6b64ef55e2
-size 8923419
+oid sha256:0c199b02d6d86ca8cec7ea7a4266fb74d6de1124b13acb3a73f8b62f8c4435ec
+size 8936107
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_geco.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_geco.parquet
index 533422ebce01e76e54e9820dc8936a3a4fd0eeb4..5726a234b3e831d51202a564ec05f042810c803c 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_geco.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4789af84970f8e2f1c5a9eecce96241c245eed506d61ea070cc800bbae722812
-size 75189518
+oid sha256:210e0b5c509a076cdf4b80245f52c50e86c61a7ca4fdb1029ede6bfae4ce91f8
+size 75157419
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_meco_l2_w2.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_meco_l2_w2.parquet
index 555222d176c8c9d48ae3a3c18df6e909fcc21672..8dc218815a193585a647da8ad6cd083b95c25f75 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_meco_l2_w2.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_meco_l2_w2.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5601cabb7eeb62282eeef99faa8877f3c2c7026f84921892f829efd6ba95471c
-size 3873879
+oid sha256:b06e53062a952f333d08948e5f8be675fc42ddbecd4959dd9089a2411d97300a
+size 3393373
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_natural_stories_maze.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_natural_stories_maze.parquet
index 47b07fc0d1fdc3b5cf37099dd68e163267be188b..071dae051366784dea936082f2db1d7749111b7e 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_natural_stories_maze.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_natural_stories_maze.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:55839e9bd372349d66238692766006c63de0ba4b0f29cb34811aaed3c13fb9fe
-size 14509749
+oid sha256:c15ea1f43c7baf47b99b4d70f6127392b8417f629f0248dcf5c4310ab607da4b
+size 14516390
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_onestop.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_onestop.parquet
index 54663e9c1d689f20f5dbbabd304a17a513330a60..93a2ddcdf4ce2cd6ffa0eaae84c20335ef726d71 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_onestop.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:61042aaee81b698b730ffd30575a1fe3ed19718192fd7fa1e43ffd4449245048
-size 45476137
+oid sha256:39004a9b95d1af5b192c708020c57906c2e747d7264ee5385364c500027043e7
+size 46815074
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_provo.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_provo.parquet
index 08b378ea408770d2eb3f557868aebe6972868d61..2541be384cc0a01d3fdcf5f9118d8a44f876c20e 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_provo.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d3ab26628a8a2148b0e4c3e1d629d1f4b25d03eb4910644f680a784f8ccd8870
-size 4047649
+oid sha256:3961b39054cbfbd67dc6e22866fb98d6a33c6bdee7d3846c53e6525c37a94d47
+size 4021089
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_sbsat.parquet
index 3089fdded50ed4c5eaedc9a2e1e40a2310b8d862..e69bae622ef394b49d6f25ee1842ffbed3761a33 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_sbsat.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b32675d6d8146abe95960dc02140db88e27cfe9f19c1e433d21b31a91aae1ee3
-size 3729880
+oid sha256:b56bfe1264bb669c0a540e2016f2108009795b0372a9f6da2ac332e85b35e54c
+size 3709724
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_story_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..53f60c87873ea12751e7c0a71695ea19b4c8f63c
--- /dev/null
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bf6f69652d6e191245fba0c0fe04eaca23ecbe49dcaf65fe9954d6cd0ceacbb
+size 3464981
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_federmeier2022_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_federmeier2022_n400.parquet
index 698260be2da7fdab4a5778d7d73d3a7ac0acd3e9..76e81a80e843d08bb2203dd5d2aea6d488774d54 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_federmeier2022_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_federmeier2022_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b7f88bfeef8693bc89c27ec050fd5030eeb768400c8eefc6a8acfcbcf5f9c67c
-size 9405302
+oid sha256:0b07cc576da66f659aa7aef1acf6efcd2d7f32f41a99a036dcc4a3c80891971f
+size 9413747
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_n400.parquet
index cc66a58bd637ba42bb0768ac1b77b3852bf6dac3..14e8b008a2b09de182bf99ac4734905ef28f4875 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_szewczyk_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2dfd90a33de612b75fd6ddca7d63ff50d8ed010c899f549df1f7c42be0348ddd
-size 20785978
+oid sha256:1fad306ecb1c6097a21fa871d8a102a8579bea6922efcbe9f921da4bdc780923
+size 20799713
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_wlotko_federmeier2012_n400.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_wlotko_federmeier2012_n400.parquet
index 6f8e30c097d21aa53f26d6f6f2e0b82da41418b1..4522b57971a583385f5fde04b5284c3a6d630830 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_wlotko_federmeier2012_n400.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_wlotko_federmeier2012_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6c6289ca54f42e3abcf7f70ad106274a525fb7366848658011f7fd87769207de
-size 5101121
+oid sha256:431a607ffe224b60b1abf523099174650c522850d6a398d2d1fa64b922c1f897
+size 5106032
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_zuco1.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_zuco1.parquet
index 7bcd8ffc5f312a10ce671b77131fd45f12e349ff..01b208471dcfcde14312169c4de8976b89030411 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_zuco1.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_wsg_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fef99304c471b1b20ccdc537fb74778f38550d9da765556552a4e1fe1d74e0f9
-size 9133424
+oid sha256:cbe25aac950e920e544e9129cb1d81adfc1d13de9ef8d88a0787d5f4d313c420
+size 9150461
diff --git a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_zuco1.parquet b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_zuco1.parquet
index 282c18b54dd0b9299fe066163e7a9b6c8bd26ccd..bdbe78b095540f909d1066c901859a18c216e872 100644
--- a/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_zuco1.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N512_M20_a1_s0_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0a41c18fb9578a8de99028cd8a793ae15e543410645164fe6214d0b472f4f5a0
-size 9071707
+oid sha256:780000d490be45f4bda2f4a635b3d4ce8cb64f67aa8bb7f01c84aca1438f63cc
+size 9077180
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_celer.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_celer.parquet
index 1a1a3e5dda81d5e6261868521ad5e6f1be47e299..1762b71c63723956480b6fdfa96a2a79b7a14091 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_celer.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a0d0a6bdc530930ba87c27b456ab568921419a4841646c71130a717406f7ad13
-size 63296463
+oid sha256:b679da27d48c11c6ed8a028940ea23021020bf510751b576270848b309203916
+size 63292403
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_dundee.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_dundee.parquet
index 63c5e6fbb80213009167c3c089015519582f1562..09dd9e1df691eb24257c25131544bd0e6c11ba56 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_dundee.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:372027ae51c3e57de0fb0be958c53a1d712879c0ff88acbe07cf70b5f3f0bd27
-size 9879724
+oid sha256:bd62fa78ecd3780be0aa2fc6a6c8f2089c0d0e317501e7aadf38b5c6ef8f1340
+size 9922004
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_geco.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_geco.parquet
index 5c7c3771a71bffbc1a653ffb5078b1b7d76154e0..0c39edb399a687143d7afd194c08a1e7e3c2062a 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_geco.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e4f60eb533a28cc407992c08bcb4e914ea24bcdf7a1e981cd9ab66e9dbc07627
-size 11229036
+oid sha256:3dc39e128c67d8244cda859314e9027354002f14f737142390670c0ce756be6e
+size 11212362
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_provo.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_provo.parquet
index 52564ea1299fc9f03789890d7c66397c1aa072c0..2b56037a935b98497df07831e241b93e81ab12ad 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_provo.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:616f5ca357a103a8e434e81e72e54c4378e117bb8579506ac1e0be0aa283cb79
-size 688575
+oid sha256:502392ba5b472d0c2487f2393ac82b439e6e781342ce946121af52754738e98d
+size 684742
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_sbsat.parquet
index 8b0b87788248e5bbd1775ba7ae0cc48ec11bbd32..f09155f3a37147d2ef39eb0d2aa57de84e3102e5 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_sbsat.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cd3c9caeeba91854e0f91fce1229a1b9e088122d7f62dc6976772479c4541e92
-size 625841
+oid sha256:21bf9a94f63c798a44fd8bcdb97b611f202110c468c03064ee5d2c46d5ae0fd1
+size 628428
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..a6d738fd73ff7ef64a33d34a8fd4c75c12468c56
--- /dev/null
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49bb13244af8f5642c2dd15ea62680f0680ce4a594cfe3c4cc23a44478acf174
+size 598973
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_celer.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_celer.parquet
index c6658ca687d1a8482dfa3d154e2125983eab1475..b2cfc5c3b90ab6183660f56aefb0d97510f0c2f4 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_celer.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7aef3bcb85265477ad4d25521ebe93210acbe78ce3a984efec931c5f3b736072
-size 63672483
+oid sha256:68f66a2ca26cd827929dc8f6f178c7df170411ac0ab2f03155dfdfe97876f79b
+size 63665554
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_dundee.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_dundee.parquet
index 78a094580a30492ea0ea3f7875069f14125abd3e..74f411b12a3cdd0ef133b32a7a1b739616877f16 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_dundee.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d711619ff1228fc23cfb08d56f85c0a6c34c0cb1de09d262b75a2417fe68cb9d
-size 3321126
+oid sha256:312e199b0df7a1ffe90e9c076bf8007ff1ce9015e91b1ab59f5457688feb269d
+size 9930138
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_geco.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_geco.parquet
index 02821b6ba817f01b74b8564f6b879f9a442703f9..31acb5338fc4dc1ab8e0f1fc551651dde9c9a2eb 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_geco.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0d7ce5c62af8d04fff26a1e0aa3f6e604df47c8ad0eb2b251b1ffddc095d2f77
-size 11280937
+oid sha256:e60f940dd33240296432d620b604bf6a4b0c7bf12a1079801bda799d760992ab
+size 11278535
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_provo.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_provo.parquet
index bc5870a8f2414c4e15ef2ac210c3768982d55bc3..319591a5dc46cb84f36f351ed8101f7268475067 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_provo.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:414f4776173cda43fa3deb912caa3f83b110d4a8b5d0bda1299c6e34b28dfe67
-size 697231
+oid sha256:80e103a526e6963eaa68ff7699dccaa274e9552a40ef1c7a46ce879687264402
+size 695502
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_sbsat.parquet
index c5bde5cbfdb49428addb9a138ce2e5ea63f0d9e3..872b6ffeb3bac8d7e7b2ee01dd9c630ed82ebb61 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_sbsat.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1cd389ac3a8b5c7dad4f04c3f41948ba15f0ee3cdbcda4caf1aa436829b6a439
-size 631503
+oid sha256:f67c6d6e6bad56385257cf337371b89ab282a2d4bd874ac7bf8570e718b62751
+size 631657
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_story_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..a8995270d8eb56863fe00223b781a71543c8afbb
--- /dev/null
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a0.5_s0_wsg_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bc1d965012167330d5cd28c7df0357da323f17941fca34a0922070ab7990a04
+size 603166
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_celer.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_celer.parquet
index f5c2c126c22f28be7a86005a5a05919d86e8c859..17aea8c97ad0107c62a13a5379af7a194f19ddac 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_celer.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ffe884561b0fb452d8253a8f6c321b4762bb00175de95b800f30808ab55ab74b
-size 63296389
+oid sha256:476b57945f58578aa3bde190452fea567f7d6c6914211e9c4d1a5537616bd35b
+size 63292329
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_dundee.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_dundee.parquet
index d485e9c9ebbfc719159bb7b198609d7176990ce8..a2f39dab626a7443eb91fcaffa451751a0e3b61d 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_dundee.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dc8779a1de5fa93acbac922b4b793d486b1b9dd0e2754b56b472b7e6ab5d380e
-size 9879706
+oid sha256:09c922e90d1c7beccaaaea618e951bd7257da0f982d93a051077df5ed649c2b9
+size 9921986
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_geco.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_geco.parquet
index 7cf446dc035c4f6f81fc43625a94331c13bff3e7..311923613a7e3ed8fc2a225363c4c3a011724e75 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_geco.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2d583aa2a90525885ec2e4fb713ca8d4e9d186b53b24e0487392e0bfaa3ab406
-size 11229018
+oid sha256:6a905869a327d466a8736dc2dfb7fa3c60f1703a9a3c97a2a33d1ca7abeab5e7
+size 11212344
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_provo.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_provo.parquet
index ac9e42e344ee0007b4ca744a90a4c3eb85bc3c25..7a8acc6bfb5bf5ccc16c433ad932a55125a038cc 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_provo.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9d2b5b133a272b6ffef4c169725aa53f34df0136b55a3557abf0e9ec94c4c454
-size 688565
+oid sha256:427b53ced3d53b0ffb2e4a1476a6eae30b8fa2aa3484bc46bc76a56fe2ccecbc
+size 684732
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_sbsat.parquet
index 99b42873720956ca653f693bc4d045b37392af2b..806749c2e746f9abb79a475e987c14d4be71cc44 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_sbsat.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:deefd40f0dfd85e236d39e4980b803370a179d061232a226c221d1c8f7fd38a6
-size 625831
+oid sha256:58c03396bcfa5f914f9a774a3b175fff7ebe1f3d09adab40226b18fc626f3f29
+size 628418
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_story_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..23b1508b23ec69a18c1e00f1768eb99570ba07c3
--- /dev/null
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e82ce11910148251769e9122a75043909740a70e1981c0992f5e0517f236571
+size 598147
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_celer.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_celer.parquet
index 8a74c1fb6b1f397242debcf19f9f013ffa5a5727..920a26e0c3ccacfbb3c6de1b9932641d6d7c4593 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_celer.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e2ed91b759d6e81c3df7ad566693d6d066fea9e82bd628106d7d161f98f71b22
-size 63672409
+oid sha256:5baa8245b79cc285db1bcaf010fb2c272e4b363ac73ffe64c61cebe53f974131
+size 63665480
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_dundee.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_dundee.parquet
index d9a0fbd7c87ed8fd58909faed987d2cd7f25844a..bc8bcd6b8e5fa6cdffefff871973707fb93b23f3 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_dundee.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4179efe10a50b3d24fef34c6ce9a75fbfe1f27cc4f71eb2af437ace6adbc629e
-size 3321108
+oid sha256:0a3eb195bfe717c01a7e1d9cb60d22ae707d82b418831087906bdf7da43356fb
+size 9930120
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_geco.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_geco.parquet
index ac97058bbae2d95c170ab23074f688c879632d3a..0acef58b4993987eb585ee8ba025c24619adfedb 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_geco.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6a018084ef4f82b0907ff03f2877bf26b04e61afc57a82c5f66e6bbac1b62d9e
-size 11280919
+oid sha256:b8264b926cd922d10f9b9c29b90340c0a0c09b216e6bc1a42a47b498b1e4ac5e
+size 11278517
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_provo.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_provo.parquet
index a53ce9133f27a8caa2c3d2c60cc6eb26a5eecf43..c6f04957dee02b0142f593f6b91845ba821da71d 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_provo.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:234e12eb092b69003cde880014826a8f968a9fd649bfaa7d2cbe8362c1ee1865
-size 697221
+oid sha256:738ade236fefa252b7f3d5d3bd74f03ad94984813f4d6c53db2c7fa874c7082d
+size 695492
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_sbsat.parquet
index afffb1e9e5e8f3f478b881f801ae329f0e212e3b..058c047da3c1595e3c036ed530bb2b88b870b692 100644
--- a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_sbsat.parquet
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7493d25498bfbbb70512004e33a5306e1cca9ee3dbe7c5991b2eef27fb632db2
-size 631493
+oid sha256:86f942e2a0e64e4fcf267ead867758d0476befd9d7e3bf46e8f87226f343bf6c
+size 634217
diff --git a/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_story_sbsat.parquet b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..33c46115c4dbbe4bb81a6ff307c5baaaa4ac9ae3
--- /dev/null
+++ b/mcword_samples/gpt2_mcword_entropy_N64_M20_a1_s0_wsg_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e8e8a86122290a976a3b04c12fd4717d8ea386a4973662d405cc1d162baa1fc
+size 603155
diff --git a/meco_de.parquet b/meco_de.parquet
index 6d016c603ca7a97e867321f455ff8a4d1478ed2c..c3f92f674eae8bc70627b490cbd7463f5e67ea49 100644
--- a/meco_de.parquet
+++ b/meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:336700fb04967c2c2bcff6ab1c0e687f49ebbcede9db0ac8b182e6aa37824dcf
-size 945171
+oid sha256:253942e992648149f7087d7616ad77aabec11e2152665cd4342fdcb5e362fa55
+size 1190133
diff --git a/meco_ee.parquet b/meco_ee.parquet
index 1f35add06776e815ca0bfd786f2ab24ff2649296..94c3657f503a33312182575f9790301efffc170a 100644
--- a/meco_ee.parquet
+++ b/meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a79b85c74051a6a3ac30f8883084ee701deef13130c4c9f883b5ee2daa94f209
-size 702782
+oid sha256:05981999893c59fc6462ce8518af83458426f07da552eb30270e7aa8211eb423
+size 935950
diff --git a/meco_fi.parquet b/meco_fi.parquet
index fe3836aaa0bd71b308c9c37f1daf4eaff9cf4932..44c682f431e09e74a9bb9f52a3e48350828b4a0d 100644
--- a/meco_fi.parquet
+++ b/meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f708a0c0ce1925f7ec644a7418662c93a1128efd86ec7ab8e1313c81a112e09
-size 692542
+oid sha256:27bd00286d698da87625ebab5c65d6ceb6f13268df7bafcb885f2621a88224bb
+size 986640
diff --git a/meco_gr.parquet b/meco_gr.parquet
index 020c76fa2b73db2c51c6e2dc95b8a8ea70734d8a..02ecd3b36a92bd9cea3dde070cdebccc6fe3ee2c 100644
--- a/meco_gr.parquet
+++ b/meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:48e306f31ad0051d4a3b69927bde46bc51586b6fb18ae09eecb049c09401c4cd
-size 982082
+oid sha256:bc1fd5a9a5284de9f60a197695ac44771aaa724ad252674e829a527240840e99
+size 1287723
diff --git a/meco_he.parquet b/meco_he.parquet
index 2f5213cd8cfeb27b48b082c1ad9e44ed2e7c7817..8eb6deeb9ad236a829790a6c03eaaed86196fab1 100644
--- a/meco_he.parquet
+++ b/meco_he.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6cdc2430a29b24dba3be22d461165e717249cdfc1445852d8644ac7f4819774c
-size 913607
+oid sha256:8d06c87e8ea050dce00a4c6198a3fc63d8586143faa1ab1aadab7e004018843b
+size 936941
diff --git a/meco_it.parquet b/meco_it.parquet
index 777cda81384da4134b77565dce7e3900ae748c34..8346e89e7cecda166b352b72f455b868333b5bea 100644
--- a/meco_it.parquet
+++ b/meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5a1ce344caf05c515e0eb78f068fb4781e73ab0304d9513e7c6cfa8740410333
-size 1130173
+oid sha256:76d1c909bfaf6473c7609e404fbb57a74cebeca478de9c455bced4a29a0516fd
+size 1252395
diff --git a/meco_ko.parquet b/meco_ko.parquet
index 33285404ef4dc94f5650698695cce50bb59acc37..0b255e24677fa0b8c1bf3a83e6fcbcfb24cc45ef 100644
--- a/meco_ko.parquet
+++ b/meco_ko.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b513a307879f5bf20a0c4be1c5455e4fc49b3e4f16cf6d0ceed44acc3b8df32c
-size 960748
+oid sha256:f4bc0f7cd310e8d87bebffc56033c539a8210f5955ac2e1619b324fcbf6a34f4
+size 1166021
diff --git a/meco_l1_w2_ba.parquet b/meco_l1_w2_ba.parquet
index a33eb3f3363c8e6fcf5369c267ab77a80215a8f1..fae6f4f3fc17cfe945e5b09c73f3c9304afa7366 100644
--- a/meco_l1_w2_ba.parquet
+++ b/meco_l1_w2_ba.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f6043072c648c37da1290d73503dc3d6d041303befa41a47b99d58d605bf4fd1
-size 153265
+oid sha256:4e2f12a85a9f26865419071dbb593340f844cf4cdb18411df59962f2a3c588c7
+size 239454
diff --git a/meco_l1_w2_bp.parquet b/meco_l1_w2_bp.parquet
index 8daab1a0e65c2ea02061318a5a4383f69bb717ee..0ff11c7d41836d244362888b7b4e4d81312ad9fa 100644
--- a/meco_l1_w2_bp.parquet
+++ b/meco_l1_w2_bp.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:168dfb960a842510cd9c8a86e73bf9c7f841c148beca14aec2200684424d82e4
-size 215460
+oid sha256:98f58fb1d0dd1f7042f3eebec083f75a08fac0fe434283fd7ec2176e373b5e81
+size 312562
diff --git a/meco_l1_w2_ch_s.parquet b/meco_l1_w2_ch_s.parquet
index 645155cf2132f09d725902f61322afb5c8cbf2a5..f9f34104942ea79197d6a509dc90f4ed946a16f4 100644
--- a/meco_l1_w2_ch_s.parquet
+++ b/meco_l1_w2_ch_s.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4b5f86e07435075fef649f2eb2a8d36b64753bbb8eed2cbd261cad483b339772
-size 170221
+oid sha256:d8500df3d9d79eea0b6b5c372ba309599893c3d8569b93e8bc6537d6e6833132
+size 247552
diff --git a/meco_l1_w2_ch_t.parquet b/meco_l1_w2_ch_t.parquet
index 89c6746fcb706c83a48e139da779eec7a242835a..2577b2f109c2e68d94d3d8c8e3f0cd85879a75e7 100644
--- a/meco_l1_w2_ch_t.parquet
+++ b/meco_l1_w2_ch_t.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4583694937af5dc6dbeb1c109737d0491cdb75ee88588735d36821eead6e5b84
-size 190251
+oid sha256:5deff9955b152f3bdd72d770e534f183415119d4a48df5cfb2632223d8378c56
+size 265185
diff --git a/meco_l1_w2_ge_po.parquet b/meco_l1_w2_ge_po.parquet
index 8d3e4e266f90bca6259402971966d6bc17b45258..2ff3e80411ea6f8d995a4360c5079cb8d65e0467 100644
--- a/meco_l1_w2_ge_po.parquet
+++ b/meco_l1_w2_ge_po.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3757ecbed3723cd1682f4ec9526afe36f6d4767842fedc0f942b5ca81e6461f6
-size 195084
+oid sha256:89a619b730877c74b92a11d34da0d4289754db9afca57a10b022bb9504e8fc18
+size 297604
diff --git a/meco_l1_w2_ge_zu.parquet b/meco_l1_w2_ge_zu.parquet
index 715b64e772c27ec37c8aa7af6b4c9a9aef16ffb6..bff1bd48cfa1ab5f83edc1f8d24b3732513f2295 100644
--- a/meco_l1_w2_ge_zu.parquet
+++ b/meco_l1_w2_ge_zu.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4d5743a321167881542798ca240a2b2039187c61d52421a439e0b7b20ec831c0
-size 195115
+oid sha256:d181565dd901270531826dc5c6c5db66321718a55a571da4789d51992b2502ab
+size 297604
diff --git a/meco_l1_w2_hi_iiith.parquet b/meco_l1_w2_hi_iiith.parquet
index e1607fe1de82476b56633f1b8012fe9a8e04f004..62bccb3615e9d5ebaa9b2f41a9bcb9d65d89f958 100644
--- a/meco_l1_w2_hi_iiith.parquet
+++ b/meco_l1_w2_hi_iiith.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b14a808459665d9bdf04105e4019fb341df15a08e6beb75749500772c54f4bd5
-size 203559
+oid sha256:58397552f10c02c173b858b5b7f03325d5473be2655383543f1db9e6cdf08786
+size 309956
diff --git a/meco_l1_w2_hi_iitk.parquet b/meco_l1_w2_hi_iitk.parquet
index c79d093100373b47da548627e6b1da9777c7ddf7..3f1a781902ff6fc17642d8bde80100e5a0cd1206 100644
--- a/meco_l1_w2_hi_iitk.parquet
+++ b/meco_l1_w2_hi_iitk.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54aae8bf20c9ab2a19484de7fd061869b297b97f69c128c10e357573f3d1346d
-size 182056
+oid sha256:139b982b8ed2cabfd04e7d60fa3f3f9ef262a456f72c6cf883e6d34ca1b7aab4
+size 286126
diff --git a/meco_l1_w2_ic.parquet b/meco_l1_w2_ic.parquet
index a10f65568f8e16c43d0639670689d368acbd0ccd..2841c3cd7e26d87b811bf0f0fc676310ea57dd1d 100644
--- a/meco_l1_w2_ic.parquet
+++ b/meco_l1_w2_ic.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:19a11450f95d10d8965794a704efd52b188e32e296b014b162121ddddaddb408
-size 184235
+oid sha256:f222216c621d0ea6edad2afe3a1dcdafddea8bc7a2882bc321f0bc614ebbe462
+size 253484
diff --git a/meco_l1_w2_tr.parquet b/meco_l1_w2_tr.parquet
index 0735bef67305c3a5ca7e5e6fa92ccc52e20d9e3d..1ffe70ba5a397a1d0eb6988d8392b29ff094f991 100644
--- a/meco_l1_w2_tr.parquet
+++ b/meco_l1_w2_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:17f369985d04896a60a21a3cea203525a36f3d5225711ab8328456a7959c05c4
-size 166053
+oid sha256:6edc426715d60e3cdbbdef09b080d9a7dc9caa9d0648f5e2c50a9d23aecbcd48
+size 270131
diff --git a/meco_l2_w2.parquet b/meco_l2_w2.parquet
index e72c4ae9a6dfbed1d07024216f35092a96331b52..29b1fdb0a0087ce507921af620d81581a3c82e4a 100644
--- a/meco_l2_w2.parquet
+++ b/meco_l2_w2.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b4b1eea550512da1a842acdac0de12e84a429cfc1e6269bf258ec90c0e3efdde
-size 1329496
+oid sha256:97b21c6caa1dfa9a7018ee34b192b29d900c3f17820c4812cfbc5321136ac815
+size 1386632
diff --git a/meco_ru.parquet b/meco_ru.parquet
index b8b3ce1a000cede8873b1bafeb85e00d93f94ea1..f736657223d5cb2ec876aa1e78c19a49cbcfc0b7 100644
--- a/meco_ru.parquet
+++ b/meco_ru.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3ea70671d1965abe962347d618dad855b6d1103239656caa0c9ef6259d1c2b88
-size 868156
+oid sha256:30cbc3b6e3407cf5118cd6f1992472101fc991a323043d9e35c7dabc8b325b03
+size 893224
diff --git a/meco_tr.parquet b/meco_tr.parquet
index a5cb4bf22930fefb130ddbe380b7efdc74b8d7ee..971b8cedee77a9a89e27950f34f5b1b39d2ad0c3 100644
--- a/meco_tr.parquet
+++ b/meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3da2e6d3d22c808c7a3e720ea9e3da10e1fb6c40f92a0f2f4ab466997a2792c9
-size 946405
+oid sha256:b24bcc61727d471858e6dd8f45ea541ede992093607cd6c36cc17e3d6e1569ee
+size 1097094
diff --git a/natural_stories.parquet b/natural_stories.parquet
index c81bb4a4b2ef79337c1eafdc9c65dda00b1d8123..42ead343f8d293ddfe7b026111be6e85486013cb 100644
--- a/natural_stories.parquet
+++ b/natural_stories.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c248902c340ac713982f35d6e47b11b40da4aeb05e40bbe60e24f27222971dbc
-size 8134665
+oid sha256:7181cd761a15868a5154e2e2a2ddf1569dcb01d48d0d8536ed11d4b84542151a
+size 9064134
diff --git a/natural_stories_maze.parquet b/natural_stories_maze.parquet
index 71a7cf0a726611e475a80d579b4b2a5ce54055a9..5e9373d01698f24781dd03a64cc718a538cf6a3c 100644
--- a/natural_stories_maze.parquet
+++ b/natural_stories_maze.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ef337252ec5c15ddb711a10a9edc628954cc37b8108d2c0961fa89d1ca24d554
-size 4394143
+oid sha256:3d56abb2b0a45ae13f11f7d07dfc6cf2c21662d81b4930e677ac11f9250f5cb7
+size 4990536
diff --git a/next_log_probs/ai-forever_mGPT_ft_entropy_meco_de.parquet b/next_log_probs/ai-forever_mGPT_ft_entropy_meco_de.parquet
index 0cd4b0ac3d410912b39ca51c6ff6f3b81d4fd7c8..8966517d6fcff917a2c5c6cdd46328c85b3e6417 100644
--- a/next_log_probs/ai-forever_mGPT_ft_entropy_meco_de.parquet
+++ b/next_log_probs/ai-forever_mGPT_ft_entropy_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bab21434aa4944958702bc03ae06b9fd9101cccd525676981cb6e41c0892bbd7
-size 548724231
+oid sha256:faa0af5a90ee1ff85936e6decd17d963e79c23757bef98a9a8c17cff011983ec
+size 547214886
diff --git a/next_log_probs/ai-forever_mGPT_ft_entropy_meco_ee.parquet b/next_log_probs/ai-forever_mGPT_ft_entropy_meco_ee.parquet
index 3b7d2df66c1c863201c50a17aff57f25ebd7251c..104743f596604b507ea5e8665dbcadd36c6775bf 100644
--- a/next_log_probs/ai-forever_mGPT_ft_entropy_meco_ee.parquet
+++ b/next_log_probs/ai-forever_mGPT_ft_entropy_meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:36e97b416bdaf86b15eaf658444d7048d1fac445c75fac6054970d48db7d5694
-size 415380067
+oid sha256:1c59ccdbc27a82139d3f5200190c0072bbe7ad56b5180f5aa22282189f65b8f2
+size 414639388
diff --git a/next_log_probs/ai-forever_mGPT_ft_entropy_meco_fi.parquet b/next_log_probs/ai-forever_mGPT_ft_entropy_meco_fi.parquet
index 7681787f164e68d3025aa9d64764671967e7206c..724d8e66274b89ab4641742e11ba8fb51b51556e 100644
--- a/next_log_probs/ai-forever_mGPT_ft_entropy_meco_fi.parquet
+++ b/next_log_probs/ai-forever_mGPT_ft_entropy_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3b488b0427d48c9e24077ce7071cef86f741305b7aceef17a8497476a74d3688
-size 400733857
+oid sha256:238ab59140b989cb545d42d843f4b8d04fbf809bd6fd002848c798368f3193ec
+size 400000663
diff --git a/next_log_probs/ai-forever_mGPT_ft_entropy_meco_gr.parquet b/next_log_probs/ai-forever_mGPT_ft_entropy_meco_gr.parquet
index 463b1ae30ae15d96a9a2b884847f50b8b21ba02c..7beaf4bf8081193744f202084bfe216851588a86 100644
--- a/next_log_probs/ai-forever_mGPT_ft_entropy_meco_gr.parquet
+++ b/next_log_probs/ai-forever_mGPT_ft_entropy_meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a7a07c059c595fa9fd908367c10469a7c11b28833806e0bdaf312aa67877ff38
-size 558791664
+oid sha256:c8e13965d3c63859a83714d161c0beb4fd3766f403971a740b9ee6d2adacaa20
+size 557339726
diff --git a/next_log_probs/ai-forever_mGPT_ft_entropy_provo.parquet b/next_log_probs/ai-forever_mGPT_ft_entropy_provo.parquet
index e6c0c9fb2c132b237d6557b3aeeedc99fc119ecf..457790d9d5529b402854268b19fb539c897ff8c3 100644
--- a/next_log_probs/ai-forever_mGPT_ft_entropy_provo.parquet
+++ b/next_log_probs/ai-forever_mGPT_ft_entropy_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a41b79b5be1bc5e2a8846ec5856354463f69a68357861770dd2bbb36b1e3a45d
-size 724278567
+oid sha256:7a40ba7b60d5790ff7359fe55ac5c38a06e689a790a11006f580cf2b0286efff
+size 737991728
diff --git a/next_log_probs/gpt2_ft_entropy_natural_stories.parquet b/next_log_probs/gpt2_ft_entropy_natural_stories.parquet
index 9fdfcb3ef4082d29d3a80891ff1dbb3caec52558..f12060f33ae6621fb1f23d68ac82839618181796 100644
--- a/next_log_probs/gpt2_ft_entropy_natural_stories.parquet
+++ b/next_log_probs/gpt2_ft_entropy_natural_stories.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7303ebecf9ec6c96960690ac0124b1c699fb1c94134e5183b26b1726d9c202d9
-size 1125547622
+oid sha256:2cd1cc546c0a68298cc6104e6aec2e7c72db37aaaa4af0ca7d0ec4cb2531c979
+size 1122855536
diff --git a/next_log_probs/gpt2_ft_entropy_provo.parquet b/next_log_probs/gpt2_ft_entropy_provo.parquet
index cb7cb732b338d3937793e17167346a1075fe1082..6408dabc049b963407169543fca55eac194382c4 100644
--- a/next_log_probs/gpt2_ft_entropy_provo.parquet
+++ b/next_log_probs/gpt2_ft_entropy_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:38c4d40d1c7c53ad28eec42e365b33c10db7e209de734d85cbfb6ee13c5324ed
-size 296120938
+oid sha256:8f5909fd876f94a23bdd54c8d1b53253c07980de50b700c6a0eb5dd30c1350b8
+size 301278343
diff --git a/onestop.parquet b/onestop.parquet
index a8399d640935c158fa2af6a5d3116d123a07df61..5996bbd7e384363ffbfa338ba69a2ba496d708e3 100644
--- a/onestop.parquet
+++ b/onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:722df395bb037fe49f7622a6f5b5567c4a4ea93a64b58e8d92294763ca490577
-size 22541856
+oid sha256:319adbba366627305a4d5a83bf1b8d6171238a3caa59774adae9bdbaf03dd686
+size 25977821
diff --git a/potec.parquet b/potec.parquet
index ce73ad39919f77bcd57fba785e044d5b3328ee45..57ade6fdc5705593a861840fed250d28affb7840 100644
--- a/potec.parquet
+++ b/potec.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d79914b7546b047abcb4072ff1eb7bc3a176068b30cf8bd4e3d7c1ab9a6e1d4b
-size 406066
+oid sha256:ebb4c60b9745e0d0aa685553628fb623f1f1292760c417285463bca8e5b63d76
+size 581470
diff --git a/provo.parquet b/provo.parquet
index 5144338d21314a67e48f915364e9e28d823eb5a8..df3f2bde17b0614ac5617df2bfb7aa3e84f9cfba 100644
--- a/provo.parquet
+++ b/provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d97872cf433d3f7540996450163082f39239f523c00cb738edccb2469065310
-size 2521468
+oid sha256:21d5436a6a9c55b6e7196941aceaa591e85356660dfe0b4ace26eef1b02c7f61
+size 2742604
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_celer.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_celer.parquet
index f6faf6f339384080a046b436a60768475285c1cb..722edb73742b6850060bf95dab1415540af50437 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_celer.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:775b450d695fb5f8f15fc2eddc2e0503b61d3a0dd479b8904e8f7a11f6a7797f
-size 4614372
+oid sha256:674e90bff678ae4314e3ec0a46b0837a7248a26368529ae5470343c522e24554
+size 4314278
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_copco.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_copco.parquet
index 59ca3b42df941ec8b00791fc3f73d207715d6d9e..51697ec8a39ede869dd54ebdc90270c26eef1cbc 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_copco.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_copco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d04a771acb94844ae002c68a350166d845556fb2fe59c66496c2d3844990a040
-size 532152
+oid sha256:33faa1af870430956c76f2bdeb7d268d022a9fe0b6187085399a307660f2fde2
+size 534505
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_dundee.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_dundee.parquet
index 19b2c827c642371ad43fd6b0aa358f29662afe69..e859c9087da16aa9f765e12c93df22b3b72315db 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_dundee.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8d17bc3709538a479fad43188ac4b0f80462b1ba37d1c86a24a609cb525ea938
-size 910253
+oid sha256:b4351ab8f3666e934ccde95b1b8d60acf874c00e498f169fadd7a7b0db818414
+size 929217
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_geco.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_geco.parquet
index 71acccb98388217fbaa99741e3a346ae72c3b48c..6fd84ba5fd673cdf13549b9cb415d366981d2009 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_geco.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0afd43bd4ff1b9392145c33496c1edad70dc5ea11cd021fda6a7c73c084504c5
-size 931277
+oid sha256:9c2f799c1e18d452695772e1d90572bf45f263aaeaeea4c475511420ad072570
+size 931588
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_de.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_de.parquet
index 1ad008df4295cb47bc2ae2c0bb4e9191f4c7153f..9fd50a8cb8bcd60982bd79827ff70323c0967b7f 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_de.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5b4c1ce1c4ffcd56cb3a15111cbac02fbdadc5c1d60a97a553d08819a7372da9
-size 38160
+oid sha256:2dce35e4a5ac3bfda124d66c83230137f87992ea4a5a06e6dbac9d9bf06acbe7
+size 38942
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_ee.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_ee.parquet
index e6a59a28bb9cd7f60eed70fe0c8cd00b307ca91b..52c922a054f2dfeb1917306755dc45715238b6d1 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_ee.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6884e2eb1db1e04204fb24789cff0c374e696b25daec282b55eb09fab31edff0
-size 30434
+oid sha256:ca7cf99c5a1a3f56508d849536445635b40dd86377eca7fbed890f09dc1ae77a
+size 31299
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_fi.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_fi.parquet
index 34d9cdc5fbb74738b6e9a498a293e961bf4ff2cb..ec8c3268fcbb5664ef95465057baf43165f7bf96 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_fi.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:77f1ee1a6afbf613751f222ec704dfbc3c0eaeb17fc6ae430962d1a6c905ea95
-size 30556
+oid sha256:c67826934ea89080e4408fbb685911cf062a7f1514e56b627576d035f77d0db2
+size 31313
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_gr.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_gr.parquet
index e7be937997054e2cb5c7e16784bce25773902a18..69ed5a427ff1ee1e40287819d82675b4ea9a6c68 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_gr.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:068fc28482a42e42cfc865c2aa93727de11330a15bf0b10205a4e2c02a8663de
-size 40996
+oid sha256:5d7583092ed0c663ecf65f682bba795827984a05bd1a974e6ebec49ae28e056e
+size 41964
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_it.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_it.parquet
index 30168922b57ea8785956629bd1a617c573639de0..8135378ee2052a2e8d1f8e97ec6cd589d2170619 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_it.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:86acb2a6d29043e1b7a7d08cfb06718cc5790ae87af80c09a6c36a34e7f93588
-size 39228
+oid sha256:84777cf3979bb6810db72717e5670d8a42a00194847b3c6bc79b3c95ef312322
+size 40167
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_ko.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_ko.parquet
index 7c7199ad760b571e3e107548e66354125251fbd0..c8852f39d2d195bbccf76cd787b1176cc4141e19 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_ko.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_ko.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f2e3bdfebe65dddf61089c026020b55a278db1d1e0447a3e0b12586aff67908a
-size 37042
+oid sha256:e3bdba5347ba0f6df49e6ff317c1d82a59752021d1d729fb4bc3938ee82dcd4a
+size 37866
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_tr.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_tr.parquet
index 1947099bfe55b82d787acbf0a7bd740eb5aadac5..2099a829a5262efb89ea10a53265615b16220371 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_tr.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e30f928283522438d8ecae5249f8156d86cdad0dbcb0b6341a1fbd728b237b8c
-size 33901
+oid sha256:aa5a047718e0968f217905d63dbacef7991a4cd362d5f580a548be9167fa93a6
+size 34657
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_natural_stories.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_natural_stories.parquet
index a6e39952cb99b513600744152500833be419cb1e..13ad7813b335c680a8a57121669474976052dcef 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_natural_stories.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_natural_stories.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4f9e36ea4da4c3f4820affee091e781c342f54c0bcc4abfface3e04a14b8e7f7
-size 175469
+oid sha256:099feb528e89b1c1f76244a35eea5f37e14022f67623172d045d122bf38ee2a9
+size 176341
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_onestop.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_onestop.parquet
index b6acf906aca7af77853c33d87fa83e4e458fa038..3ba4808cca5618278aed993d588296da5d0868ad 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_onestop.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ca8bcff55c32038570ef56a767ae2440cd9e6d9590acc0a2520a7d4e3bd2b356
-size 580988
+oid sha256:fcba8fe4fb54f8a5407f91e417036792a99df2ff8acac65004061413833eb303
+size 592354
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_potec.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_potec.parquet
index 7de53537b432eb0ae8d07aab0bcb75000a112acd..9e0f615e3557e5d62f8a6137d131b3dc31dc0a15 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_potec.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_potec.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eb65862fe9503e8ef4bfcceb11f69ba3d1b85c71fcea9253bfb054a991180f15
-size 35308
+oid sha256:ddfe8213faaf21f2c59260fbd5e76596869f0a7fb841f5e76cdcd14a103dd028
+size 36135
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_provo.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_provo.parquet
index 32d93d5d4a84c451b236e70839f067cf361b1857..197ebc1d23b01b776cfa155f56e3eb43a92fba5c 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_provo.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9cd6843665491883e582b3c610de60b129a7ae6483fac509bee099eba51ee8da
-size 47369
+oid sha256:a50e54841d47a73884c19ec14218749acabbaa77d517e7137107d0fe7232b0d4
+size 49234
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_sbsat.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_sbsat.parquet
index 9d466cce16cb14ecf91032ea63d917e0a7266dc7..1500134bda630c82264753144b53b80af4be15f3 100644
--- a/responsivity/ai-forever_mGPT_responsivity_N512_s0_sbsat.parquet
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1a19acd552f0fa3975c090431eaf8d10ac836327001ebaddcfa7460fdcb75834
-size 43777
+oid sha256:0890737acc74f93448ea9000766f10ff8fde9c5d25f958b78e5395fd918bf482
+size 44558
diff --git a/responsivity/ai-forever_mGPT_responsivity_N512_s0_story_sbsat.parquet b/responsivity/ai-forever_mGPT_responsivity_N512_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..34bce81062eea76c8ee5e70ddc97e5c1710a9c1b
--- /dev/null
+++ b/responsivity/ai-forever_mGPT_responsivity_N512_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7175cb5a1b3751f6c5510960c0c8b1906c1216996b01cbaa48365c5e6ac569fc
+size 44766
diff --git a/responsivity/gpt2_responsivity_N512_s0_celer.parquet b/responsivity/gpt2_responsivity_N512_s0_celer.parquet
index e8571c83c24a7119bd1f2fb505a02bc0b83723ed..630a15b76174d91d63aa5cce2fe0e10b3f8101ac 100644
--- a/responsivity/gpt2_responsivity_N512_s0_celer.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b4b358407e151a1e96ecffe99f504b254382225cc098a959f354e86216628ddd
-size 4732942
+oid sha256:da03a3df25df9d6dbf0dd242ea8fc652080004199ce1b6b9a59860044d723227
+size 4409618
diff --git a/responsivity/gpt2_responsivity_N512_s0_copco.parquet b/responsivity/gpt2_responsivity_N512_s0_copco.parquet
index cd4b20bb7410e6d8c29795ffd79225a9c406031d..1612cd719a5bbd4971ccbcd3ebf3a007aff3528e 100644
--- a/responsivity/gpt2_responsivity_N512_s0_copco.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_copco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aa741bc0d72a56e833d04c57e5276cebc36d9d2dfe911d7b5efc7810d59d0a36
-size 524595
+oid sha256:ed15b7616955cea0bd9e774ef99619ae927ef8ccaafed1c5c8e0e610d869aa45
+size 525939
diff --git a/responsivity/gpt2_responsivity_N512_s0_dundee.parquet b/responsivity/gpt2_responsivity_N512_s0_dundee.parquet
index 2b1c2fc1384906e2f5c1fb955557af3fac070b0e..8fbcfe7b84c6f43c5b09d32ad3ab1c46809d87d7 100644
--- a/responsivity/gpt2_responsivity_N512_s0_dundee.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:30b7c421c0d65c1d6932455c0e9df5a6260bfff155a5b039ccc1c9d3d032118a
-size 925927
+oid sha256:b4b65a4e582bf12c63a4af5ae74c45a14b4793244f67332cfae9684bba81995c
+size 943771
diff --git a/responsivity/gpt2_responsivity_N512_s0_geco.parquet b/responsivity/gpt2_responsivity_N512_s0_geco.parquet
index a634e29bcecde11533ac0cd0d3ef766fec8b2954..4c3af2fdc033a55606300637185ef54ceb7d4ba5 100644
--- a/responsivity/gpt2_responsivity_N512_s0_geco.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:474a92e48e6717b4ccb07a3348d3718b499332cbde68ddad0f08cc9051dcdfe0
-size 949506
+oid sha256:a195f49a319a210fce31624071609c5dc02aa0ab0b57d9cd6aa056ab4afc31b1
+size 952290
diff --git a/responsivity/gpt2_responsivity_N512_s0_meco_de.parquet b/responsivity/gpt2_responsivity_N512_s0_meco_de.parquet
index ca6d6ba00ac319c4ffc89359cb60b1f644b9cb7a..490bc53e9909bde30a5a32e6cd7d38a33f531494 100644
--- a/responsivity/gpt2_responsivity_N512_s0_meco_de.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_meco_de.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c1bd3c22bb163ada38045acc0093061e8dfaa886fc6bda18469195b2bb6398df
-size 37334
+oid sha256:8dfb89b002696eb50e5233ddcbff52fa012326cce46e066aa51418c1bae7b921
+size 38167
diff --git a/responsivity/gpt2_responsivity_N512_s0_meco_ee.parquet b/responsivity/gpt2_responsivity_N512_s0_meco_ee.parquet
index 7508726e2d8b726c0df96c880ded819bbd43156e..a8813a367f71aff770d1546210868d6db2c7de8e 100644
--- a/responsivity/gpt2_responsivity_N512_s0_meco_ee.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_meco_ee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a155491ee775f41e765c2a34d4201e95c7d81c787d6fd0c2c021c755de181950
-size 31063
+oid sha256:590d60f1fb90690c7756f0a3c120ec952e60469bd26cc1b9fd3c970ed97b7937
+size 31776
diff --git a/responsivity/gpt2_responsivity_N512_s0_meco_fi.parquet b/responsivity/gpt2_responsivity_N512_s0_meco_fi.parquet
index 42e9f9d561389c5d901b9b69cba0bcb54036d062..617d3688fa2ceac1f9917d9c52478edc75a7f126 100644
--- a/responsivity/gpt2_responsivity_N512_s0_meco_fi.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_meco_fi.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5db63766795ece610e20636908630db9ddc5f7277d5b81a7fc6ea843d590f11b
-size 30548
+oid sha256:c1ae44521da8ecbd6c4baad19e9d380c678cfaa579a0c7187aa8eedc338b01f0
+size 31355
diff --git a/responsivity/gpt2_responsivity_N512_s0_meco_gr.parquet b/responsivity/gpt2_responsivity_N512_s0_meco_gr.parquet
index 3ff88c64ff53fcad04bb63948ad6cbe26d0578d9..68465a7a983ef99c6537c604be49f7d8a73edca8 100644
--- a/responsivity/gpt2_responsivity_N512_s0_meco_gr.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_meco_gr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9639c2dc8216f1865c154d455db3a1cbd0171ce209544801224770ca947eaed6
-size 40827
+oid sha256:cf851c30ed812609260e433873982cb700f2801c38edf9a5c5a36c01ff321a6a
+size 41431
diff --git a/responsivity/gpt2_responsivity_N512_s0_meco_he.parquet b/responsivity/gpt2_responsivity_N512_s0_meco_he.parquet
index 81961af016679a25ec956a2e2fb0a51f2e47d949..957be9fad88fdf41f9e44253d696d117a857ecc5 100644
--- a/responsivity/gpt2_responsivity_N512_s0_meco_he.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_meco_he.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:36582fb952ca6665269581af21662fbf19bdf1ac142fb5b56b3e6b8a760477f3
-size 38430
+oid sha256:1948fcba2e0f2a5066a285254833d1de46f924ea50e6b16c4461b8960f06479f
+size 39330
diff --git a/responsivity/gpt2_responsivity_N512_s0_meco_it.parquet b/responsivity/gpt2_responsivity_N512_s0_meco_it.parquet
index 711b3ee455b580e5fdc5bbaacdcab15a5c9181b2..bf7a8721e3a2113827e55cd364cb9c5c0de5ddc4 100644
--- a/responsivity/gpt2_responsivity_N512_s0_meco_it.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_meco_it.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1422d65b517036ee738e75bf1a884a613c9d030de07b866ca54836ca9c992deb
-size 38386
+oid sha256:664098dde73523dbfffaf153c69ad957ce080546edb7fb9a9d445c03c3bc3ec1
+size 39081
diff --git a/responsivity/gpt2_responsivity_N512_s0_meco_ko.parquet b/responsivity/gpt2_responsivity_N512_s0_meco_ko.parquet
index 37de9e5a1fcda24a7352c0066cde9e20fe2cb499..84e8bb983c055186d93085fed749ac6c94aca27c 100644
--- a/responsivity/gpt2_responsivity_N512_s0_meco_ko.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_meco_ko.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0029a552fdc02b946cbb2848e6697503a3391519817c9a3c9d717515f0e3d806
-size 37237
+oid sha256:b20cc9cbbbfaa9afb3ad3c395394e1fb427d20767f74a03550f260415d073f4d
+size 38046
diff --git a/responsivity/gpt2_responsivity_N512_s0_meco_ru.parquet b/responsivity/gpt2_responsivity_N512_s0_meco_ru.parquet
index 19ea99ae67803525717c2c58cc8babea2174594e..8c247e22b20605787b10b24f45e0520af660fedc 100644
--- a/responsivity/gpt2_responsivity_N512_s0_meco_ru.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_meco_ru.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5eb99d200fd84acae646640143ef480254097cd0bcf7c82dfa92920bfa9acfe8
-size 38108
+oid sha256:ff9627a906aca87df704bff3df28d4b868ffecdad913a8e36b3cd091a7ef798d
+size 38815
diff --git a/responsivity/gpt2_responsivity_N512_s0_meco_tr.parquet b/responsivity/gpt2_responsivity_N512_s0_meco_tr.parquet
index 5c243b346b78862e6c1792cf49801c9caa47a2f8..8cc2e25c6911b67d7a036401cccd29b0cff69d27 100644
--- a/responsivity/gpt2_responsivity_N512_s0_meco_tr.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_meco_tr.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:45f1c15f3046f360585ed74853bd3f2b985667812388ba23940242930f48e413
-size 33699
+oid sha256:29b8e2fd488c9542585bf0279478f70b346a3487437c2eac3146358602754fed
+size 34309
diff --git a/responsivity/gpt2_responsivity_N512_s0_natural_stories.parquet b/responsivity/gpt2_responsivity_N512_s0_natural_stories.parquet
index 3b05a077126a45d7c33711f8beb2bd176171ddf1..81601e1f3aae372bc9c1c933bc17aa1d7cb261ae 100644
--- a/responsivity/gpt2_responsivity_N512_s0_natural_stories.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_natural_stories.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:16d52bf2a77c1625dd1a400f52929e35c4be3a5097f3e0b37f2e80cf40c7b728
-size 177291
+oid sha256:2ab2ce1bef11a57a9e2456576aa45b7d33f767ae317b16c72b16632cedc630cb
+size 177998
diff --git a/responsivity/gpt2_responsivity_N512_s0_onestop.parquet b/responsivity/gpt2_responsivity_N512_s0_onestop.parquet
index 4fdf21ed8012fd266770841b3e3b22fd634882be..4d08e3481aae830ba1cb087f9ce002cfdc71b778 100644
--- a/responsivity/gpt2_responsivity_N512_s0_onestop.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:55776d2b410f90e9b2b5ce2e6424c30e8ca72020cc8ad38c2b3517f74db5d49e
-size 592107
+oid sha256:68234f11836dddcfdb9894a6d8603397d9ba4d80b2f955073f87aa59110619fd
+size 602543
diff --git a/responsivity/gpt2_responsivity_N512_s0_potec.parquet b/responsivity/gpt2_responsivity_N512_s0_potec.parquet
index f33f39b128f2ecdae5a5323130f9cb99277aaa44..f687f04e2b09da4b9840762fa12bd15cb5c39ba4 100644
--- a/responsivity/gpt2_responsivity_N512_s0_potec.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_potec.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3b6e6df71beee6811d5427ba0abff77f01779e34efe330b4bdf5706e9c0ffd18
-size 34579
+oid sha256:17df0cd6337a617df4e633de49b0f19c9a0a873fc3e7417934ecaaa457217526
+size 35347
diff --git a/responsivity/gpt2_responsivity_N512_s0_provo.parquet b/responsivity/gpt2_responsivity_N512_s0_provo.parquet
index 562fe920646d0e48f442f2151c9858bdf8ec50f6..e217716c491e44b303439b45498ce9eed9a9b817 100644
--- a/responsivity/gpt2_responsivity_N512_s0_provo.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c1aa883fed4910afae9c12ecfed467e8c78e99caf159095d23a81cbcc978cf0f
-size 48141
+oid sha256:2a54812a137acd67430eac5f9bc92066eda1b6907fdd3a9372baab7ab8c33b78
+size 49785
diff --git a/responsivity/gpt2_responsivity_N512_s0_sbsat.parquet b/responsivity/gpt2_responsivity_N512_s0_sbsat.parquet
index 2b05e3ed85a7a58a5592665acd516219a8cda45c..25116815655a7bf23542d9abcda955784c997101 100644
--- a/responsivity/gpt2_responsivity_N512_s0_sbsat.parquet
+++ b/responsivity/gpt2_responsivity_N512_s0_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7f675d282bfe0fcfd3a60f701d99b0303f206b8c20cf5e8de545025faf45494f
-size 44407
+oid sha256:08312091f5eb8a7b25aa908b79ca55d64909466487377c18d8d1d31fd83647de
+size 45266
diff --git a/responsivity/gpt2_responsivity_N512_s0_story_sbsat.parquet b/responsivity/gpt2_responsivity_N512_s0_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..ab7feaf1e93312e3069d6507a98a2d5996364e68
--- /dev/null
+++ b/responsivity/gpt2_responsivity_N512_s0_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:070f03c0fb2a90a180828ba5f73c7a2b8247bffa8a0d8cfafd4bf331a5b11cd4
+size 45330
diff --git a/sbsat.parquet b/sbsat.parquet
index 84fbf368b43d5266e0b5bcbb646c44dc7eb9407a..3583a61b462b53b7800f96961085e912af52ce86 100644
--- a/sbsat.parquet
+++ b/sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:56b1120b064f60856debc2649bff6d5c08247429b6514c27f6ff4bc5fc445e2f
-size 2135117
+oid sha256:76693d3e1d7959200edf92e3c9a714591078f29f9318d1bc1812597aa7ca4895
+size 3491927
diff --git a/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_natural_stories.parquet b/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_natural_stories.parquet
index 04b275de4d257ad83185d4be3b1ea20b11fc4843..232946b5973a80d74090fe7632879ab3a986af28 100644
--- a/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_natural_stories.parquet
+++ b/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_natural_stories.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:46080eddab67fa630b6f149088474dca735ff0b75d801e2283f3ffb259c65bac
-size 1063748
+oid sha256:bc4c422739a59f0bb1b1851b0ce803625f4e179c832fd46c8e5cdb5c7cb1dea0
+size 1063906
diff --git a/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_provo.parquet b/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_provo.parquet
index 0482a54078c57c310fa54ace5c3392a404f848f0..0f6f6221602b3614bc5b33e38f11cce91e25e1e7 100644
--- a/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_provo.parquet
+++ b/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:611e527f81cc7de0087f818046ee9d3cda6e8dcd814060e105d563bd170da1a5
-size 273921
+oid sha256:1746e3e040b7d018d7228bbe6a5049a0d23159d657651630941a3a9b9e1f2122
+size 277348
diff --git a/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_tokens_natural_stories.parquet b/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_tokens_natural_stories.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..68c95d33e8248e8ae4a7eb88c7c72e236f60a09d
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_tokens_natural_stories.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc391b716c9881f8b67dd2d2b9698f46c67416dd3d3aa716300864fa788233e5
+size 1174578
diff --git a/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_tokens_provo.parquet b/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_tokens_provo.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..12023ee662d67b7a0d32d3a6e39ef11269fb023c
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_contextual_N50_s0_tokens_provo.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bff1b506b2551849f8a15fd822003a1e0ed6499f427c1604cdc447094cd0823d
+size 295055
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_celer.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_celer.parquet
index c3e10eb8875388b4a8f48470770c6e97edff4a6e..c836097476b8f4f4109e0194b849b6ca24f06968 100644
--- a/simadj_esim/gpt2_simadj_surprisal_noncontextual_celer.parquet
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_celer.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9eb42ba8c0edd944597fd15323438bbe8b56556d74a0f795af476154421cc110
-size 29126187
+oid sha256:a3a4fd5721edec74c203bffab817290200925644ac9ea01da0216ee8146a4af8
+size 29364983
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_dundee.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_dundee.parquet
index 967ae4041ce4aa7610de40bf391a91a3ddbbec48..3697ab9a7df033d19b9b139256c2a8e2a911c578 100644
--- a/simadj_esim/gpt2_simadj_surprisal_noncontextual_dundee.parquet
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_dundee.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d1ef1971a8530ea454cab1338bfc1920f594add6840f67fb4c3fd1fca2e5b28d
-size 4701697
+oid sha256:dad4ab48ac3578d9ede4d93a56beb2c7b5561a2b02273dbf343d3e2966b4db87
+size 4762065
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_geco.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_geco.parquet
index 3c5fffd4821f3e95e74111418bb23bb49baeff6e..f156134c27d98a4dd916709286b63c85154c798c 100644
--- a/simadj_esim/gpt2_simadj_surprisal_noncontextual_geco.parquet
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_geco.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:38d0c925bebe9f53e1275bd8834e318c11f7796af05a4a659baae8ae0b010706
-size 5684323
+oid sha256:e6cc4b470d567452c35742ccbcb75bb2ad15b67f383a636c911acd9312b1989f
+size 5680026
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_meco_l2_w2.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_meco_l2_w2.parquet
index b70f29f452efde0f1ae3c184e7e4b1e4ee9f9474..0ede1ef9b0444d39e4b26e7aa49eb9fd915c26f9 100644
--- a/simadj_esim/gpt2_simadj_surprisal_noncontextual_meco_l2_w2.parquet
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_meco_l2_w2.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:03b039a05d148d44008a9bee611f6c1f713e8f6377fea9da6878611885a91c48
-size 246128
+oid sha256:e6c0eb8d6f686cd4a79a8110b4a83f1625ac48269dc78cf916dc1aa81551d5b7
+size 192824
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_natural_stories.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_natural_stories.parquet
index 92e58b640f6f6f2afc20696a90064746e87af2da..162e709f9aa4481aefe9f290c04c65096d7178b9 100644
--- a/simadj_esim/gpt2_simadj_surprisal_noncontextual_natural_stories.parquet
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_natural_stories.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f411e458d96ff9d20eac5a9b930d9a2a0ae2b8f92de6dce220055929d87c7201
-size 1112952
+oid sha256:063b939d037f93ce3554e83fcb7728eb898a71c669bf802b65e74bcadc94d89f
+size 1112944
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_onestop.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_onestop.parquet
index 4aebddea9cb63f43237cfe0043252ffa3f445ad1..db33804435727bad52d1aa93a5db4c620312d4f4 100644
--- a/simadj_esim/gpt2_simadj_surprisal_noncontextual_onestop.parquet
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_onestop.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:867fca4ae4191b1f2e7ead4e143fca65ae0343a7857b40cb570e981f607b932a
-size 3511842
+oid sha256:7afbbb6a25861f4183dbd82971d75b5faca4155992851455df79993e9890ca9c
+size 3588933
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_provo.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_provo.parquet
index 508b0f58f15d80b545e066af41bc19969675982f..6d8ec417aa19f0303a208c144ace2ff2c8c56e44 100644
--- a/simadj_esim/gpt2_simadj_surprisal_noncontextual_provo.parquet
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_provo.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:032501c93dc3cf1413560641b4afe514ddcc0603bdb34fb3722a32c48604d20f
-size 286648
+oid sha256:1bdb036da68912216c936c71cc8039229abdcdd3829a25a1ffb6adaf9dd52efd
+size 291251
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_sbsat.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_sbsat.parquet
index 0c3f62c32ce21ea549ca356ad17f06f7b258ab7a..d5f1811d74cb65e5f1dae1ab1414e7c0fc1be8e0 100644
--- a/simadj_esim/gpt2_simadj_surprisal_noncontextual_sbsat.parquet
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_sbsat.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5f1bafa8b909094967c487ae3b94ace37c3d9f3a60928feaf415c61b68f71470
-size 277530
+oid sha256:f620223fe50cd8703aed919c5de233d0cd88c93950e7b1f4bb59acc3daa4f8e2
+size 274920
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_story_sbsat.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..3f8f5bd6b4c59daf38d00a455f7c0dd2a2da6e74
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4c6430767f99b6fb2868df9b9a9ddfde3644ade6f162a14bdcd4736cd75bc64
+size 274271
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_celer.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_celer.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..8b0d0f9e2aaec8947610f8b889763e32a6a8ee7f
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_celer.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:273479e1d1e278e0393aebc092bae59538e0e05e506385c4528f55ca4946be79
+size 30710870
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_dundee.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_dundee.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..a73185a0b249034c35049a7093a5e178e041627e
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_dundee.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b3c51599a5f51eab9fa8ff123e8cd14c46478319f33031ba3a25497356a74b1
+size 5210309
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_geco.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_geco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..9e9676e3187aca494a5f56fd2211291d4caeff7a
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_geco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd0395261ef9ce8116c8c6cc64f17ca947319b7d7c4e29760448d34b49ee5ee7
+size 5985339
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_meco_l2_w2.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_meco_l2_w2.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..8dd5ae339c214c49e2e75ba45b5e1b82fd86c0a9
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_meco_l2_w2.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:847e963e6b18f69ad31c1014c96a49cb5769246d1595ca8e1784c013b10bcaf0
+size 208680
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_natural_stories.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_natural_stories.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..f8784b7757c1e638e87501004bdbc0fa97646c72
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_natural_stories.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c8ecc670350f0bb37b45998420149e746af56d4769ee535f94f032b8cbdf637
+size 1226551
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_onestop.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_onestop.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..9ebddecd50a7fedccac1cd28d1a72261bf6eacbc
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_onestop.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef6a4a11db8196d1eb07d6aed6bcff7c23af24a19d38b5c56b68d437c5c3e4bb
+size 3769368
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_provo.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_provo.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..472d36da251ab7d1a7d85c744805ab6dfb3d2263
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_provo.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:149950ce82d70e0dc39805fa4d7e6a61baaec4b4a91eac919ea462b1d58d5b19
+size 308969
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_sbsat.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..2fef4db70fce910afb3285ffa16b75d84bbb50db
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18d350651ce4c6590767fa9f97279d3d8f240d8e7a95bcc89e93a22c96c2f4aa
+size 293053
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_story_sbsat.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_story_sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..9f0876b5594eebff41bf1abb6fe36e991993b4ff
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_story_sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20e05bb61d53c2d461b8972362340a47e4044039323a8aac9a5057a14399dcb0
+size 292325
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_zuco1.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_zuco1.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..0067ef9cf51d215afb7f00a9fb0134a53946fac1
--- /dev/null
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_tokens_zuco1.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e82b8f58a9a98317a410f77f4d6258c741d1ae23721387176275fa565917252
+size 821952
diff --git a/simadj_esim/gpt2_simadj_surprisal_noncontextual_zuco1.parquet b/simadj_esim/gpt2_simadj_surprisal_noncontextual_zuco1.parquet
index 38e214091a668719cec1b410d882bcb8e1d54036..3ce1b78219b620d38ddd8b7ad4555c3fd992b074 100644
--- a/simadj_esim/gpt2_simadj_surprisal_noncontextual_zuco1.parquet
+++ b/simadj_esim/gpt2_simadj_surprisal_noncontextual_zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c5eec981c5c7181c13c7f7c20eb60fbe40d66453fd7b8b357f9da78d2f738175
-size 782624
+oid sha256:6040f9e0fbf80e8e5592e7a17610141323c9d08744f9cb7d3b8ffe7b045097b5
+size 782919
diff --git a/szewczyk_federmeier2022_n400.parquet b/szewczyk_federmeier2022_n400.parquet
index f6d487a26d725c6cbc96c1035a20dcb77907ce86..fee7825d45b89b4aaaceea0fa80bf0cb3dadc7ea 100644
--- a/szewczyk_federmeier2022_n400.parquet
+++ b/szewczyk_federmeier2022_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7199f14714362f4537eeb699873037ef16128f0cefd0d090c2239f873c582bc0
-size 2150638
+oid sha256:98d95e863e5c6b3d064cffbd7de5c2687f7a3cc6eedac4b1f3c90a8229791574
+size 2428112
diff --git a/szewczyk_n400.parquet b/szewczyk_n400.parquet
index 331c8f5c37591bcb9ff430758b0fa520913eac64..910f90c2000ed0cf93802ffff47d5d8eb51563ba 100644
--- a/szewczyk_n400.parquet
+++ b/szewczyk_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8e2b7d144154a58a56786f5179fdbd5f59083452ecf9551a14d7844e81c588b9
-size 4190927
+oid sha256:3d8d84b2a8c446ad98b9afbc621d62e620e616d2177f7728625ca9a9237d0eb1
+size 4512201
diff --git a/tags.parquet b/tags.parquet
index 9e810a6a7037f3c17f0bbf49652a0c5d9ceb4787..26b8cc4ec06b4e01e31d61643a419f62e5a62da8 100644
--- a/tags.parquet
+++ b/tags.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c18349948d07ea1f7a80d65e0796e352e44e6e2e36be81c5024f06eb8e9ecc0f
-size 57990
+oid sha256:41fcd4e6438f3a410ac2a131fc010e6023a9821ff7a27934f584200ba57d5b6a
+size 73464
diff --git a/tokens/celer.parquet b/tokens/celer.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..768ca93be4b3bd771d6cf2a58fd5384113e967a0
--- /dev/null
+++ b/tokens/celer.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4223c177c9e54e7e111081606f32d23d065cd44f48599e81f0b199e195421d5d
+size 568188095
diff --git a/tokens/copco.parquet b/tokens/copco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..4f103adc871d421d586317dc1134b7fcd3ffbc72
--- /dev/null
+++ b/tokens/copco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df39b9363e904d82b9580fc76b739edd8b8461800d083c0b65982da1ae7b2ce3
+size 8082596
diff --git a/tokens/dundee.parquet b/tokens/dundee.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..37c3bfd4fb9c656a365708b0235ae1100c0268b0
--- /dev/null
+++ b/tokens/dundee.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b56d6100b5efd08c8683610830d1033c10e6f3e558f543b0987840af2284e9c9
+size 101920178
diff --git a/tokens/federmeier2007_n400.parquet b/tokens/federmeier2007_n400.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..f0bb119f40b50b8d2cce0dbb301be0ab28ebce0a
--- /dev/null
+++ b/tokens/federmeier2007_n400.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f6f272a418871b53c3e5803232dca7928b4db7a6c85ad7a5b44a488454d72c6
+size 751611
diff --git a/tokens/geco.parquet b/tokens/geco.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..eeac7813ab0a48bceaf4d43d34d4a27382e6cd81
--- /dev/null
+++ b/tokens/geco.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3944be3c587cbc6627813f9153e0a4177f54ae66e7aefb0962bd65c0e6b00c2f
+size 111935300
diff --git a/tokens/meco_de.parquet b/tokens/meco_de.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..77e21bd911a661554fd8ea44560f985270b78a6f
--- /dev/null
+++ b/tokens/meco_de.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:639f0b0a2540c6f83efd6048b139e2790d9a4295d7a87793a18b4415818e4377
+size 841114
diff --git a/tokens/meco_ee.parquet b/tokens/meco_ee.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..e00f17c06c1eed9122ee6892148d657d120bce9e
--- /dev/null
+++ b/tokens/meco_ee.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92e81496158f81f8c2fd4cd03e7ba460c3b85b22617cd361d4bbc30ebd995dc7
+size 934023
diff --git a/tokens/meco_fi.parquet b/tokens/meco_fi.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..aa505704625dd9685fbeb1e43f73e2a6d6555c81
--- /dev/null
+++ b/tokens/meco_fi.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28a28e4a4f0b4f87273210d1b347abeb867cd6d6cc3c90e022da5a8051de7fd8
+size 933541
diff --git a/tokens/meco_gr.parquet b/tokens/meco_gr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..d50bf008722ace0e8cda3b7c74e480da2f4c78c8
--- /dev/null
+++ b/tokens/meco_gr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d72372f4c3b24bdf4cb1c2d42e824ac526bc9d284915eef8ebe0eb845bf4f34
+size 1217121
diff --git a/tokens/meco_he.parquet b/tokens/meco_he.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..7543bd0a8ea5671165c1cde2d5ccd0bbecb5655c
--- /dev/null
+++ b/tokens/meco_he.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6862ef8bd6c3ee7054cd76002425619fab7a832d53811d7176bc2dddacbb3394
+size 256717
diff --git a/tokens/meco_it.parquet b/tokens/meco_it.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..342a9fbbda8b728ef9e5cc7b24b2db8afb5d4e22
--- /dev/null
+++ b/tokens/meco_it.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:950a243a3cbe8c53ec7880a4240a32e6afbd0f4f9d8116d4005d695a4816417c
+size 818662
diff --git a/tokens/meco_ko.parquet b/tokens/meco_ko.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..d054b72628dd73511bf2f9278c4e70808a0a25bc
--- /dev/null
+++ b/tokens/meco_ko.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ad4ace5a3c34150597b1dc890d3bc11138bd9f72748a9e9a78068617ca9dd20c
+size 1096443
diff --git a/tokens/meco_l1_w2_ba.parquet b/tokens/meco_l1_w2_ba.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..ccc07a83d63ca2c14efde0321c8da3599dd474d0
--- /dev/null
+++ b/tokens/meco_l1_w2_ba.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:227ff6134c900747cb412c053ea3e11ad57dd96bab76596b79c82975765c0dcc
+size 343684
diff --git a/tokens/meco_l1_w2_bp.parquet b/tokens/meco_l1_w2_bp.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..f5eb61ec4aed29177a48a12b6102cdcc13af0e66
--- /dev/null
+++ b/tokens/meco_l1_w2_bp.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d73da9dc182ad096a22965a1b4c0315cb6d3908eb418458c3aa1a05be38c5449
+size 293729
diff --git a/tokens/meco_l1_w2_ch_s.parquet b/tokens/meco_l1_w2_ch_s.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..ebb571108a47f7790a022b77d4f4e4b9e753e45a
--- /dev/null
+++ b/tokens/meco_l1_w2_ch_s.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e92131712e018b0ebc5fb19f24db73916398e1a09f5337b6c89d5cebe293d8a
+size 155065
diff --git a/tokens/meco_l1_w2_ch_t.parquet b/tokens/meco_l1_w2_ch_t.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..6b1ca33869af7989e7480e3deebf3d072305c0a4
--- /dev/null
+++ b/tokens/meco_l1_w2_ch_t.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6487c9c00dca18e34b4c5d450591422c0c9d078373f073324533ed5946e4cd8
+size 153171
diff --git a/tokens/meco_l1_w2_ge_po.parquet b/tokens/meco_l1_w2_ge_po.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..8120b9403df63511f535d17ee467a961e02e8f09
--- /dev/null
+++ b/tokens/meco_l1_w2_ge_po.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18288a241beba640e7ae72dd55a45e5873abf9d548306ad59d7075cd05b45880
+size 295208
diff --git a/tokens/meco_l1_w2_ge_zu.parquet b/tokens/meco_l1_w2_ge_zu.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..3f517ea1255a63930d56d943ada4be23aa1606c0
--- /dev/null
+++ b/tokens/meco_l1_w2_ge_zu.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be03a05dbda8c480815753d0e6b0f0ebe2be240e80bf9d880e58340a2215a57e
+size 295208
diff --git a/tokens/meco_l1_w2_hi_iiith.parquet b/tokens/meco_l1_w2_hi_iiith.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..a2df8a3c4de177e704b5f1bfbfe71cfee1881f78
--- /dev/null
+++ b/tokens/meco_l1_w2_hi_iiith.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b93c01d28994e107437014b942df76351f280f3f8352bb5e0f70d73bb81b327c
+size 402999
diff --git a/tokens/meco_l1_w2_hi_iitk.parquet b/tokens/meco_l1_w2_hi_iitk.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..64b85c29a88ae9fb23d89950a84fab02de60f72c
--- /dev/null
+++ b/tokens/meco_l1_w2_hi_iitk.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5a759330b1b1cf4b410fe12b8be1887e8d1c7fabf5a43c8ef9c968ed479a21a
+size 402939
diff --git a/tokens/meco_l1_w2_ic.parquet b/tokens/meco_l1_w2_ic.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..94fa21f70ea70b9607ab483ad6ed6874d1b7c029
--- /dev/null
+++ b/tokens/meco_l1_w2_ic.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:732b63c0bf8a3f3b58484a08891e6fbbac9bd031f0406cf26f59433e06d8f660
+size 341976
diff --git a/tokens/meco_l1_w2_tr.parquet b/tokens/meco_l1_w2_tr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..244f7cc66adeb1c33dec3f56a27dbeb86a817703
--- /dev/null
+++ b/tokens/meco_l1_w2_tr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e6365e32d4c4eec6384a8477ee1189a7c5db5bc230cd4c10733485ca51c11bb
+size 333331
diff --git a/tokens/meco_l2_w2.parquet b/tokens/meco_l2_w2.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..23e7ff2da17092feeb74630ed4bf5fc5ec042661
--- /dev/null
+++ b/tokens/meco_l2_w2.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9de0c49813d130e758af76d1e7138e9903d6576d54549b86b3444ac466b3ed37
+size 469150
diff --git a/tokens/meco_ru.parquet b/tokens/meco_ru.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..682aa7ccf37c2ad6f36f7fadc621e5b56ee0ea78
--- /dev/null
+++ b/tokens/meco_ru.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d154c4a3ce7cac67c662f9dc81859df7e2f58d754cf95aeb289e155e87dc4df7
+size 291827
diff --git a/tokens/meco_tr.parquet b/tokens/meco_tr.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..f3572efcb3ad7dc6e2601e3411eadcc3763d324d
--- /dev/null
+++ b/tokens/meco_tr.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a3f88185be21ad28016e4939057f91022d390abe27734cd9fb31753331e42c5
+size 943123
diff --git a/tokens/natural_stories.parquet b/tokens/natural_stories.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..3fb62f362cfe49e425fb44efc061177529a17c39
--- /dev/null
+++ b/tokens/natural_stories.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc426cf37320c2612cc06ab95f889e4f7541c5829c98d9529aa9b6368d5a6d0d
+size 3337203
diff --git a/tokens/natural_stories_maze.parquet b/tokens/natural_stories_maze.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..59b2f9e045d45aad29655127c9ba035210acda80
--- /dev/null
+++ b/tokens/natural_stories_maze.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88ffebd1345b9de8317e746ec456c2621c8b4d1dd436945e25879b8f0b3dfa80
+size 1558592
diff --git a/tokens/onestop.parquet b/tokens/onestop.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..68ba074417129ffc5c0321bed8917cc80144f24e
--- /dev/null
+++ b/tokens/onestop.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6aad5ada91cb8f0080dee180d18b8178bf71df35eafbbff0529c8a47ab7f464
+size 67187470
diff --git a/tokens/potec.parquet b/tokens/potec.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..51dd8bd610c5c1bf4388524995a40c198128e3a6
--- /dev/null
+++ b/tokens/potec.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3515bfe3c5af764bb0071bd21e4181ddf239ea1b0a912a8119e90766e3dd3a8
+size 585016
diff --git a/tokens/provo.parquet b/tokens/provo.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..b0c142643bcfdfd27edeac36bc8efc108e247712
--- /dev/null
+++ b/tokens/provo.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec72225d1d29f05c03ecab8e4eea848135604921c0a20e43ebcd2db6d618d5f5
+size 6480346
diff --git a/tokens/sbsat.parquet b/tokens/sbsat.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..72b5965c8202290a580f64b899b2df069dec7bcf
--- /dev/null
+++ b/tokens/sbsat.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c80d64d56bb8559b9515672a5545f6ad3a6db477e0acdc1156ad387af6f7827
+size 12091562
diff --git a/tokens/szewczyk_federmeier2022_n400.parquet b/tokens/szewczyk_federmeier2022_n400.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..7a4caa9d514ebb678e031daedfd8971f816cf23d
--- /dev/null
+++ b/tokens/szewczyk_federmeier2022_n400.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16b93d52fd22490e6ca46f5db018fc05ba268afcf05bebba1077cf63449711c7
+size 675009
diff --git a/tokens/szewczyk_n400.parquet b/tokens/szewczyk_n400.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..0a587d3ce60f19f42824f63914bbb6429f98bbbb
--- /dev/null
+++ b/tokens/szewczyk_n400.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc231aa3aa41a6561f11b756fa5c50f4cd66696214173ae03495670d6e020831
+size 1305378
diff --git a/tokens/wlotko_federmeier2012_n400.parquet b/tokens/wlotko_federmeier2012_n400.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..f48a0ec699d9671c321fccdf20351455fa25de35
--- /dev/null
+++ b/tokens/wlotko_federmeier2012_n400.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f9ac634e8664716382e70fb133d4725c09a7b37ea67ec6a8f63f34bda396693
+size 519224
diff --git a/tokens/zuco1.parquet b/tokens/zuco1.parquet
new file mode 100644
index 0000000000000000000000000000000000000000..5c5ab94628f3d07a64a65a6327f2e5556929e3d8
--- /dev/null
+++ b/tokens/zuco1.parquet
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05f78a6483763a449ffcb33fc5b3c3006e1556624266c4ae76f22cbf0b00f1d7
+size 13838671
diff --git a/verification/wave1-tokens.json b/verification/wave1-tokens.json
new file mode 100644
index 0000000000000000000000000000000000000000..484da4b3dc10059e4b7d84be453a0e2a675d4465
--- /dev/null
+++ b/verification/wave1-tokens.json
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa6c99a374e4bd32594fd1de27e1e87faaabe156f4135155c90f7dc17d0c4a3e
+size 17617847
diff --git a/verification/wave2.json b/verification/wave2.json
new file mode 100644
index 0000000000000000000000000000000000000000..2607716a4407de45cc80cb0964c5ae37c31d126c
--- /dev/null
+++ b/verification/wave2.json
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5afb4037c9054933030357ee04afab2aaa5de02d290249eadf158a1a9f938146
+size 38674652
diff --git a/wlotko_federmeier2012_n400.parquet b/wlotko_federmeier2012_n400.parquet
index d08f9629ff1fd73711f127bab20af00b2949c2ce..d9d1f1f1421d5737f7fc9bf819503265a1e60b88 100644
--- a/wlotko_federmeier2012_n400.parquet
+++ b/wlotko_federmeier2012_n400.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:784f0e8f6edc16e704ca7408646f5297fa07a62e5efde76f1f0f7d69fcbf1945
-size 1700217
+oid sha256:e14003e7d30937ba5b601aef83b9cbf037fe6f2e32295b7de00a00dbba525878
+size 1921908
diff --git a/zuco1.parquet b/zuco1.parquet
index 6319edb95c9c6bbac33341c502adbf9d7e42b03b..a43b4210daf1eb847783614324d57f25660fb16b 100644
--- a/zuco1.parquet
+++ b/zuco1.parquet
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5a6c79ce74a9c01f6670f98d788939c26581677fe597f85f41c6549d9ff994f4
-size 3402356
+oid sha256:9c5f172501dbbf4f1f4bcb575cd2d0c8035019b6fa70bc1af4e1e184b98a6d61
+size 3556452