samuki-hf commited on
Commit
5bbc55f
·
verified ·
1 Parent(s): 848ed89

Restore 80 mGPT cells with corrected start records: 66 information value cells and 14 no-BOS token tables

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. README.md +2 -2
  2. brothers_kuperberg.parquet +2 -2
  3. brown_spr.parquet +2 -2
  4. celer.parquet +2 -2
  5. coverage.parquet +2 -2
  6. devarda2023.parquet +2 -2
  7. dundee.parquet +2 -2
  8. emtec.parquet +2 -2
  9. federmeier2007_n400.parquet +2 -2
  10. geco.parquet +2 -2
  11. hubbard2019_n400.parquet +2 -2
  12. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_brothers_kuperberg.json +95 -0
  13. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_brown_spr.json +95 -0
  14. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_devarda2023.json +108 -0
  15. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_dundee.json +95 -0
  16. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_emtec.json +95 -0
  17. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_federmeier2007_n400.json +95 -0
  18. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_geco.json +95 -0
  19. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_hubbard2019_n400.json +95 -0
  20. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_meco_en.json +95 -0
  21. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_meco_l1_w2_en_uk.json +95 -0
  22. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_meco_l2_w1.json +95 -0
  23. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_meco_l2_w2.json +95 -0
  24. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_michaelov_n400.json +95 -0
  25. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_natural_stories.json +95 -0
  26. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_natural_stories_maze.json +95 -0
  27. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_onestop.json +108 -0
  28. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_provo.json +95 -0
  29. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_sbsat.json +95 -0
  30. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk2022_n400.json +95 -0
  31. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk_federmeier2022_n400.json +95 -0
  32. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk_n400.json +95 -0
  33. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_ucl_et.json +108 -0
  34. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_ucl_spr.json +109 -0
  35. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_wlotko_federmeier2012_n400.json +95 -0
  36. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_zuco1.json +109 -0
  37. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_zuco2.json +108 -0
  38. manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_zuco_n400.json +109 -0
  39. manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_de.json +93 -0
  40. manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_ee.json +93 -0
  41. manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_fi.json +93 -0
  42. manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_gr.json +93 -0
  43. manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_it.json +93 -0
  44. manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_ko.json +93 -0
  45. manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_tr.json +93 -0
  46. manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_de.json +93 -0
  47. manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_ee.json +93 -0
  48. manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_fi.json +93 -0
  49. manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_gr.json +93 -0
  50. manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_it.json +93 -0
README.md CHANGED
@@ -230,8 +230,8 @@ Start from `tags.parquet`: one row per (`tag`, `predictor`, file), with the
230
  model, method, configuration, storage class, layer and provenance of every
231
  cell decoded, so a cell can be found by model, method and configuration.
232
 
233
- - 2,515 cells across 58 corpora, 234,593,637 scalar rows
234
- - 24 methods, 26 identified models and frequency tables
235
 
236
  ## Layout
237
 
 
230
  model, method, configuration, storage class, layer and provenance of every
231
  cell decoded, so a cell can be found by model, method and configuration.
232
 
233
+ - 2,595 cells across 58 corpora, 285,933,322 scalar rows
234
+ - 25 methods, 26 identified models and frequency tables
235
 
236
  ## Layout
237
 
brothers_kuperberg.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ffe406425efe3048559f6390f9c926a503d8b9382fa0ee27b10ba79e01705acc
3
- size 4159216
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:47c83b7dcc6762d7dc2b272cfea8bde468c77fbef514ab563514222b2aae4c81
3
+ size 5490871
brown_spr.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6b03f6c63d93835050aeb84e3c98678b8e35f1aa935547237b9be139056d7dcc
3
- size 4251085
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:892de7a3cb74861f871fad5b7376e3cd21e72a5d9ffb7cf72138f504014e3b11
3
+ size 5646642
celer.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:bb02ef586f7b53c1809d5c12c5a581583b3c2f296076f5ae38fde0fc03ebe54b
3
- size 142970341
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:129152a3a60e4548cef6924a076c9fd319993938402f285d7213f4fba782d0a1
3
+ size 175769998
coverage.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e49fce89d8126a019c9d69d70119f6af9051fe3f789528630fbde3b10a5a1063
3
- size 2824
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2560ba4a7a182808bd2461cd4b2fc3c23ca3340c6bee221b0a5bb0691a0aa42
3
+ size 3610
devarda2023.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8a36c2b318683c6fdbb3890e52171bc6061350cc9cb83933aac4636191a2d926
3
- size 1148127
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c654f0d5c375693ce51fb0b3acfeaccc868b358ce7bb99719ab35296c5c54ff
3
+ size 1502133
dundee.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d6fe65ac6e5a961c3af693611e87614980c6bb5a5edd86416b872169f34851cf
3
- size 34626499
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b677a607b733592a2d20811b6ab489a8fd19fe88b14b3a34c37568e2ba47e378
3
+ size 42761319
emtec.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a1acee69ac96c15e02059f951b91f59892ec7e1df05cc04da68fa2d26769dedb
3
- size 26473659
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f330f0353a63f0bd1029e06df3ef12bd9de89550baac533ea3e7911c52091f17
3
+ size 31539554
federmeier2007_n400.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:607c3efeeb557d0e4cdd3891df9858a34c814224d7d0164be38180ceb84218d9
3
- size 2024066
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ecacb3c93043e348b7ca0de1150e6d498e24251ba147c4e5f580dc4a5752614
3
+ size 2637918
geco.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8a29100fd05db02fe5889fdf51599f4e59b2f05b22e7edd1a7c3cb60f5790e56
3
- size 37563700
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef6108153fe7d7b52be19c421f19e4906c7d38983be505c6b0160d41a99e6a9c
3
+ size 46501743
hubbard2019_n400.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b31b69519c57b6f320a509ce63785f98c4c7d8670a471af0d383365711be15c1
3
- size 641061
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63b8cffd17b0c6c6c99cd8c36534702254503435c295e275635df998bd0b3ad5
3
+ size 975463
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_brothers_kuperberg.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "2ba1782989e5c3b47c4146e3cdcd1341d2318a317c7f4dbccf9667888fef76c0",
50
+ "n_stimuli": 648,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "b750299c07a2f115da1d43e5aba31ab2c1934d42df497704fa263d5e0985b8c8",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "Quadro RTX 6000",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T16:20:45+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_brown_spr.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": 32,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "240e84851c7ec7d2e0c6df9b0f46f1308da2446e62379f74b73122d73a61c153",
50
+ "n_stimuli": 450,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "6e4ce2eac4ccfefe30bd15aac645700dc385b650c44674e9afdb3ce327179aea",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "Quadro RTX 6000",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-16T18:33:16+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_devarda2023.json ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": 32,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "878f5b750b90a1027a58b6b77d10cacc9d91b0626bfc4505dc3a04c94b1f9038",
50
+ "n_stimuli": 205,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "95bec749d3c8594b7888fcb5c60937c5145e42ab74eedcd8dfdab950338bfc0a",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA GeForce RTX 3090",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-17T00:57:18+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ },
95
+ "audit": {
96
+ "hard_issues": [],
97
+ "soft_issues": [
98
+ "mean_outlier"
99
+ ],
100
+ "examples": {
101
+ "mean_outlier": [
102
+ "devarda_s133: mean=0.75 corpus_mean=0.97 |\u0394|=0.22 (>5.0sigma=0.16)"
103
+ ]
104
+ },
105
+ "failed_stimuli": 0,
106
+ "missing_stimuli": 0
107
+ }
108
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_dundee.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": 32,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "7e634bbd21030e4d5386193726ebf74957ae28f4555911a655154211cd9f8d88",
50
+ "n_stimuli": 20,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "444f11e35ee2371fbbdc8a75522fa7bd57534fdedcbe9df8e1a71c2014944bb2",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA TITAN RTX",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-17T21:15:51+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_emtec.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "c95e85cb4de0089f770bd51f83867e940b0a6c227463da3161e05fa3b9528b2e",
50
+ "n_stimuli": 588,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "23ba827729f65b8441d3efb8521276a866e712e9bcc69ded0f2781f3fa5bee87",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "Quadro RTX 6000",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T21:04:59+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_federmeier2007_n400.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "c154ff022f3a16d53313926bc1a10942bfaf5874d1da6d23f59dfd47cf1f5fbf",
50
+ "n_stimuli": 564,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "9f77b7a70d9fea6b0e3a6112062b19c7268208d85673acf7be9d80d493d3788d",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "Quadro RTX 6000",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T19:35:23+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_geco.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "29d759acef49c2eb2ab44d396f83853833e24ba5fcbaef7c50132819fc6aab95",
50
+ "n_stimuli": 588,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "a75d4be0162370063f53bca095a5b23021577af17f1d30dcffa8b249489239fd",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA TITAN RTX",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-16T06:05:19+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_hubbard2019_n400.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "251a3d9adba93f464c78a220e7970fe08c9adb0626b2bbfa5b6906e8a67b79a5",
50
+ "n_stimuli": 192,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "23c6124bb4ba8af44ca0e1bb322c56a920314a711daa32d2784d901c0262df34",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA GeForce RTX 3090",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T15:50:01+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_meco_en.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "e4e0484f2cd64ba7ef49e06d288a3bd501270c64ed75cfdb8e98a46cda9f39a7",
50
+ "n_stimuli": 12,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "c91b8f5673e30ad4237f138fd543303cc47c40698303982e28820bcc8c78aa89",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA GeForce RTX 3090",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T15:43:18+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_meco_l1_w2_en_uk.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "abd0d259f0c8170aad354dbe464d45b8acbed14e0506cd09b4c2ee1e38ddf729",
50
+ "n_stimuli": 12,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "03e81001aa80454655e78c5479c1a0efdffab92f5a16230838fb41226bb2ed2a",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "Quadro RTX 6000",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T15:54:41+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_meco_l2_w1.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "6945e14950ee0d86bf9836506ece34e0c6bafb094473362bd7d5f37b5cabdbf6",
50
+ "n_stimuli": 12,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "c638d1c67f39bbe51355f97ee77251031b9c7b61360cc15e5279176c9c5e4820",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA GeForce RTX 3090",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T15:42:25+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_meco_l2_w2.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": 32,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "4dfa0948a52136e63121cc5a0f1eec778883210edb4e1281aeddf57551e229af",
50
+ "n_stimuli": 18,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "4956892f311545a8801a5184c45ff407520852b2317306fc856c27e6b470175c",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA GeForce RTX 3090",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-17T07:22:56+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_michaelov_n400.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "6333f0f3c9fd8df8ec0988eff121b9d2d8c3afd3dd144887a7c6e349d378234f",
50
+ "n_stimuli": 500,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "0802f16acb944126df405b6694fa6daa5c64e5d71507709196a3e8f4eb7e40eb",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "Quadro RTX 6000",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T16:15:27+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_natural_stories.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": 32,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "402be7c2cf67b5b42ae803fb8da78ce93932dd082ec8e1d76abbca9e70000054",
50
+ "n_stimuli": 10,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "92e31d23dc701e1abdbf06b82e6258586c02bf275aaae0d72cafcfa508cdfce7",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA TITAN RTX",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-17T17:17:00+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_natural_stories_maze.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "732587d774fdcee65c62dd2928303306ba5d4e39adfac5a13a9e1d0e28461a80",
50
+ "n_stimuli": 474,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "528a12edff3e4665f1d9f71633a04adbf5a6cb339c546b75cc21429621c7b949",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA TITAN RTX",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T19:35:44+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_onestop.json ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": 32,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "304cd0369bc96952403c2fbf6a01bad2a8f045b24f82bbcf6309f1eab78aab8d",
50
+ "n_stimuli": 339,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "0a172a20208e6bc7019f8a50cdc967028394337e87cffe745e8fddaf0304c143",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA GeForce RTX 3090",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-17T16:01:51+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ },
95
+ "audit": {
96
+ "hard_issues": [],
97
+ "soft_issues": [
98
+ "mean_outlier"
99
+ ],
100
+ "examples": {
101
+ "mean_outlier": [
102
+ "onestop_b3_a1_Adv_p3_v2: mean=0.09 corpus_mean=0.69 |\u0394|=0.59 (>5.0sigma=0.59)"
103
+ ]
104
+ },
105
+ "failed_stimuli": 0,
106
+ "missing_stimuli": 0
107
+ }
108
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_provo.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "2e84d19d8d281c7337c5d39839bad1072854fd00cd94caf737dbbcc01caa398e",
50
+ "n_stimuli": 55,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "e012c21227916a9f39d487619ffb4aa59fa2f010000dbb53d567f91b46c64aaf",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "Quadro RTX 6000",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T19:13:01+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_sbsat.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "e56a90ffc7972e9a30a1e1cbd8fa867f1a3143564e2c2eefde2619427db53406",
50
+ "n_stimuli": 22,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "f285bca8d3555c39e0611cb2f0925ebbd38cf2885587a660131b476ac1938a23",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "Quadro RTX 6000",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T19:14:36+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk2022_n400.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "ef1222c02ddf4493a56339967dfc80ae60a327e8e72e7568cd9bc8ad077df1eb",
50
+ "n_stimuli": 672,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "a83bde83214f6c15b66425bfaaac20d2b24818bce5bb2b6928c1237553985118",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA GeForce RTX 3090",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T16:12:25+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk_federmeier2022_n400.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "30ad8dad6afdbe389fb6f013c3605ea5a54c758a67248fcadde46f9d285999f7",
50
+ "n_stimuli": 600,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "01cdebc8c32df77615e4902f368675d43502217896bf367cfc1a6a388e1ea518",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA TITAN RTX",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T19:43:42+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_szewczyk_n400.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "3310e0ab23d763a8845582e6037a0f7cc8c045bf03766db292ca0fa5af04b078",
50
+ "n_stimuli": 1331,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "24f33002b7b626ccbe89307135d46e57472036b91757bbcb8226ba973161bdf8",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA TITAN RTX",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T20:36:29+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_ucl_et.json ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "4230549facfefa797285013ca093badd0cb09ac92b146bee7d26dbef9f8b9dd5",
50
+ "n_stimuli": 205,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "3e86a607cc28583307eb7155c8c0771fd2ed3ff623b6efc8506e6ed0d37c17f5",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "Quadro RTX 6000",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T15:55:01+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ },
95
+ "audit": {
96
+ "hard_issues": [],
97
+ "soft_issues": [
98
+ "mean_outlier"
99
+ ],
100
+ "examples": {
101
+ "mean_outlier": [
102
+ "ucl_133: mean=0.77 corpus_mean=0.98 |\u0394|=0.21 (>5.0sigma=0.15)"
103
+ ]
104
+ },
105
+ "failed_stimuli": 0,
106
+ "missing_stimuli": 0
107
+ }
108
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_ucl_spr.json ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "182e70372dbd841647bd1de838bbc7126213e7d4e393b052250c42cdd829bf8c",
50
+ "n_stimuli": 361,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "19324401e60d85f4e3940a71539e16ba57f8669587ec8744b83bf0d95a308ea7",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "Quadro RTX 6000",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T16:01:19+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ },
95
+ "audit": {
96
+ "hard_issues": [],
97
+ "soft_issues": [
98
+ "mean_outlier"
99
+ ],
100
+ "examples": {
101
+ "mean_outlier": [
102
+ "ucl_133: mean=0.77 corpus_mean=0.98 |\u0394|=0.21 (>5.0sigma=0.15)",
103
+ "ucl_348: mean=0.78 corpus_mean=0.98 |\u0394|=0.20 (>5.0sigma=0.15)"
104
+ ]
105
+ },
106
+ "failed_stimuli": 0,
107
+ "missing_stimuli": 0
108
+ }
109
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_wlotko_federmeier2012_n400.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "e981a08acd68ff9644cd52c73850edf5a399e94bbf2c7bae78971273f5b91bed",
50
+ "n_stimuli": 300,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "869bc987a4e9d4271f33dcdb5f776d6847fe31b8a6b50e703cc2da5ffc5ac59e",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA TITAN RTX",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T19:22:35+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ }
95
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_zuco1.json ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "9e7ebd9f28d8a551d27c8f6211ee6e2d8dc34b56203219aa2a2962c95f6fe89a",
50
+ "n_stimuli": 300,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "f22227ee924163198802cbed3cec9c5c1202741261f37f757b8083b5841bc829",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA TITAN RTX",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T19:29:31+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ },
95
+ "audit": {
96
+ "hard_issues": [],
97
+ "soft_issues": [
98
+ "mean_outlier"
99
+ ],
100
+ "examples": {
101
+ "mean_outlier": [
102
+ "zuco1_NR_77: mean=0.83 corpus_mean=1.01 |\u0394|=0.18 (>5.0sigma=0.16)",
103
+ "zuco1_NR_186: mean=0.78 corpus_mean=1.01 |\u0394|=0.23 (>5.0sigma=0.16)"
104
+ ]
105
+ },
106
+ "failed_stimuli": 0,
107
+ "missing_stimuli": 0
108
+ }
109
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_zuco2.json ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "16b5c0cdc2945851c528e0dd9c5142a8f4a529ef37ca1a433c6fdd2f0234d767",
50
+ "n_stimuli": 349,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "7e37b923b6b93623b225ed8fbea974d68f0cd27133c5fc728ff793418de83c68",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "Quadro RTX 6000",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T16:18:07+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ },
95
+ "audit": {
96
+ "hard_issues": [],
97
+ "soft_issues": [
98
+ "mean_outlier"
99
+ ],
100
+ "examples": {
101
+ "mean_outlier": [
102
+ "zuco2_NR_253: mean=0.85 corpus_mean=1.01 |\u0394|=0.16 (>5.0sigma=0.14)"
103
+ ]
104
+ },
105
+ "failed_stimuli": 0,
106
+ "missing_stimuli": 0
107
+ }
108
+ }
manifests/ai-forever_mGPT_infoval_N100_M60_p0.9_s0_zuco_n400.json ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "information_value",
5
+ "storage_class": "unit_bound",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 100,
11
+ "max_tokens": 60,
12
+ "alpha": 1.0,
13
+ "seed": 0,
14
+ "add_bos": true,
15
+ "dataset_language": "en",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "words",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "basis": "recipe",
42
+ "add_special_tokens": true,
43
+ "empty_context_text": " ",
44
+ "empty_context_ids": [
45
+ 227
46
+ ]
47
+ }
48
+ },
49
+ "corpus_sha256": "49c92b8e7420acc4299301b0fcb6cffb4a7dd86ff17d7425d91db0030e4b9854",
50
+ "n_stimuli": 300,
51
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
52
+ "versions": {
53
+ "python": "3.12.13",
54
+ "replica": "absent",
55
+ "numpy": "2.4.4",
56
+ "torch": "2.11.0",
57
+ "transformers": "4.57.6",
58
+ "tokenizers": "0.22.2",
59
+ "wordfreq": "3.1.1",
60
+ "wordsprobability": "0.17",
61
+ "genlm-bytes": "0.2.0"
62
+ },
63
+ "unit_rule": "spacy_sentence",
64
+ "unit_layer": "sentences"
65
+ },
66
+ "status": "completed",
67
+ "payload_sha256": "428f5e8c49ca6fb1f6f65c193b0a728891ae860148a8ec1cc9532a73f80d5144",
68
+ "environment": {
69
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
70
+ "code_dirty": false,
71
+ "device": "NVIDIA TITAN RTX",
72
+ "model_device": "cuda",
73
+ "cuda": "13.0",
74
+ "computed_at": "2026-10-15T16:02:24+00:00",
75
+ "metadata_correction": {
76
+ "at": "2026-10-31T13:00:14+00:00",
77
+ "field": "identity.model.start",
78
+ "from": {
79
+ "ids": [
80
+ 5
81
+ ],
82
+ "tokens": [
83
+ "<|endoftext|>"
84
+ ],
85
+ "basis": "override"
86
+ },
87
+ "reason": "the record named the model's default start token; this cell's recipe uses none"
88
+ },
89
+ "dependencies": {
90
+ "spacy": "3.8.14",
91
+ "en_core_web_sm": "3.8.0",
92
+ "sentence-transformers/all-distilroberta-v1": "842eaed40bee4d61673a81c92d5689a8fed7a09f"
93
+ }
94
+ },
95
+ "audit": {
96
+ "hard_issues": [],
97
+ "soft_issues": [
98
+ "mean_outlier"
99
+ ],
100
+ "examples": {
101
+ "mean_outlier": [
102
+ "1_77: mean=0.83 corpus_mean=1.01 |\u0394|=0.18 (>5.0sigma=0.17)",
103
+ "1_186: mean=0.78 corpus_mean=1.01 |\u0394|=0.23 (>5.0sigma=0.17)"
104
+ ]
105
+ },
106
+ "failed_stimuli": 0,
107
+ "missing_stimuli": 0
108
+ }
109
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_de.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "de",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "ce4b1a41117e4957f9e892838fb10de008a5ab5b678487b6892424858c5d92d9",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "d16d5b3c0acf8185698416d10f5544208c875a17bc4bdf5a6420c384ba04706b",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "NVIDIA TITAN RTX",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:12:59+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_ee.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "et",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "0e862f2c2aee076712ef53ecdfd5b8e32e38fa2122af234f6438499bcffd9177",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "b8488b58e735403eb8e62a6e71bbc3ea152fb9085abe62d75864411df30d15e1",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "NVIDIA TITAN RTX",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:23:54+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_fi.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "fi",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "b5f84dde69a87b18b91479857b2c366f6c7807c3e65900cf17628016cc30a6b8",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "7f0b0fb1a24c38b07a8ab8c2073e65c7673891afdb84d650a4799984c7523ebd",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "NVIDIA GeForce RTX 2080 Ti",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:32:23+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_gr.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "el",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "bb46487d3d7a77783fe5fe5c1499a0b8098eead45b44fae5e3c113196e174992",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "159f5e1c1c528a97ed3e28a34b1e46c180b8afcfa3184b4c695d3ecab176216a",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "NVIDIA TITAN RTX",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:36:48+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_it.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "it",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "ba0323f83b8740883fff22fa2bda31381049edf8b2a483234ce4fb0194f15c7a",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "118097e7a112f9c6097a4817452e6ddf7047be11ffc73dddf1e8373c4948094e",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "NVIDIA TITAN RTX",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:13:01+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_ko.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "ko",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "75b94d84721674538046087d46de51f3faa9a774c6cd8681dd5c04fd1ef0cde5",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "447d9c94ecfa8395c3a5ec4a9f5fe8756872a24c62de073b48b8906fb1a67835",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "NVIDIA TITAN RTX",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:23:54+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_meco_tr.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "tr",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "stimulus",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "08d1818b972b20a84fda05050d019a9c651df865b00443da344e1d6e59a1a0c6",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "f6e4ebcce01ce897cf85ba30a05a7554f003894d2e2a6480fef8c10db5f9e184",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "NVIDIA GeForce RTX 2080 Ti",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:32:56+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_de.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "de",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "sentence",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "8a91f15c9c999ca3d2b81d5f0f9033275054cffbf4dac78e99339230ee70057d",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "beac50fa0c7f7bababea37555e8be3d90171e85c0df18a278785e4f900da0f0b",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "NVIDIA TITAN RTX",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:37:36+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_ee.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "et",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "sentence",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "a3bd7f10daded68cea3728334385d0638ded96ba61df00cc85eb4c98848aaaf5",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "974ddf651d97eadc7753877219960f64fd02ad8f14ad252261d6d27abf2924e1",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "NVIDIA TITAN RTX",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:13:04+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_fi.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "fi",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "sentence",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "4ea11b1a9e4318635b18087b3844bd3cfa3891f932742f2a3e93c0887c79b57a",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "793bdf6ff79d8905f1198003ee2004f5c2e23da8fb4ea197c7629c7d497140c8",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "Quadro RTX 6000",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:23:58+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_gr.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "el",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "sentence",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "3fea08cc811f5cd24cbe536ebc4360c1a6dabad4819c7515a4cd6b68ca8155d2",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "e5e920fc0576efc1406a7f6fc6584caba65f0ee442cb02aaa99df0e619c1c2d6",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "NVIDIA GeForce RTX 2080 Ti",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:32:54+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }
manifests/ai-forever_mGPT_surprisal_tokens_nobos_sent_meco_it.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "schema_version": 1,
4
+ "method": "surprisal",
5
+ "storage_class": "positional",
6
+ "layer": 1,
7
+ "options": {
8
+ "model_id": "ai-forever/mGPT",
9
+ "beam_width": 5,
10
+ "n_samples": 64,
11
+ "max_tokens": 10,
12
+ "alpha": 1.0,
13
+ "seed": null,
14
+ "add_bos": false,
15
+ "dataset_language": "it",
16
+ "language_override": null,
17
+ "chunk_size": null,
18
+ "word_starters": [],
19
+ "top_p": 0.9,
20
+ "context_scope": "sentence",
21
+ "unit": "tokens",
22
+ "window_policy": "reference",
23
+ "segmentation": "whitespace",
24
+ "model_revision": null,
25
+ "store_distribution": false,
26
+ "model_identity": null
27
+ },
28
+ "model": {
29
+ "source": "ai-forever/mGPT",
30
+ "revision": "40897bd7c8b47a76802c411108ca6220438b8b40",
31
+ "sha256": "36faadfd737c580e4d1d02f21e6b9ae4a429237a23d34b558694c9a059494c04",
32
+ "files": [
33
+ "config.json",
34
+ "merges.txt",
35
+ "pytorch_model.bin",
36
+ "special_tokens_map.json",
37
+ "tokenizer_config.json",
38
+ "vocab.json"
39
+ ],
40
+ "start": {
41
+ "ids": [],
42
+ "tokens": [],
43
+ "basis": "disabled"
44
+ }
45
+ },
46
+ "corpus_sha256": "6da2884eee23d85230abb29366689af0289e5e99fc41fecba935ccd3ed55b148",
47
+ "n_stimuli": 12,
48
+ "source_sha256": "89260a2fb0825fc83a6d71aa62274306d4cdc92da5fd703bfde302786758643c",
49
+ "versions": {
50
+ "python": "3.12.13",
51
+ "replica": "absent",
52
+ "numpy": "2.4.4",
53
+ "torch": "2.11.0",
54
+ "transformers": "4.57.6",
55
+ "tokenizers": "0.22.2",
56
+ "wordfreq": "3.1.1",
57
+ "wordsprobability": "0.17",
58
+ "genlm-bytes": "0.2.0"
59
+ },
60
+ "units": {
61
+ "log_p_observed": "nats"
62
+ }
63
+ },
64
+ "status": "completed",
65
+ "payload_sha256": "02a6bed9cc1fc92de0529d128cf895be9288bf14848fb7d378a461dc7217b654",
66
+ "environment": {
67
+ "code_commit": "d47b6eaebd1b2706bd4cbad0d57aa3d217dfd1d2",
68
+ "code_dirty": false,
69
+ "device": "NVIDIA TITAN RTX",
70
+ "model_device": "cuda:0",
71
+ "cuda": "13.0",
72
+ "computed_at": "2026-10-15T15:37:35+00:00",
73
+ "metadata_correction": {
74
+ "at": "2026-10-31T13:00:14+00:00",
75
+ "field": "identity.model.start",
76
+ "from": {
77
+ "ids": [
78
+ 5
79
+ ],
80
+ "tokens": [
81
+ "<|endoftext|>"
82
+ ],
83
+ "basis": "override"
84
+ },
85
+ "reason": "the record named the model's default start token; this cell scores without one (add_bos=False)"
86
+ }
87
+ },
88
+ "window": {
89
+ "rule": "sliding",
90
+ "max_positions": 512,
91
+ "stride": 256
92
+ }
93
+ }