Card and selection file after the cross-check: the 4:1 run at epoch 1,250, clearer provenance

fixed_split_ckpt_plan.json is the file of contrast_experiment/w9/eval/ of the code repository (commit f9c4b8b): the 4:1 gallery-aimed run is named at epoch 1,250, the checkpoint the paper reads, with epoch 900 under `previous`; three runs gain the `extended` note.

README.md

@@ -24,13 +24,13 @@ reviews (Qwen3-Embedding-0.6B, 1,024-d) and an MLP maps the pooled result to one
 
 ## Contents
 
-265 checkpoints, about 580 MiB, and four JSON files that list every one of them:
+266 checkpoints, about 580 MiB, and four JSON files that list every one of them:
 
 | Listed in | Checkpoints | What they are |
 |---|---|---|
-| `rank_ckpt_plan.json` | 157 | the five-fold runs the paper's numbers read, each at the epoch the validation rank rule of Section 5 selects; one of them is the DINO-style control of Table A1, trained on fold 0 only |
-| `i2bce_sel_pool.json` | 5 | I + SimCLR on the five folds (Figure 8), selected by the same rule |
-| `fixed_split_ckpt_plan.json` | 68 | the fixed-split runs: 15 runs at their selected epochs, one of them with an earlier checkpoint kept beside the selected one, and 52 more checkpoints of two runs that the paper reads across their epochs (field `trajectory`) |
+| `rank_ckpt_plan.json` | 157 | the five-fold runs behind the paper's numbers, each at the epoch the validation rank rule of Section 5 selects. One is the DINO-style control of Table A1, trained on fold 0 only; five, CE capped at 2,048 sentences, are the partners of I-CE runs the paper prints and are not printed themselves |
+| `i2bce_sel_pool.json` | 5 | I + SimCLR on the five folds (Figure 8, Appendix L), selected by the same rule |
+| `fixed_split_ckpt_plan.json` | 69 | the fixed-split runs: 15 runs at their selected epochs, two of them with an earlier checkpoint kept beside the selected one, and 52 more checkpoints of two runs that the paper reads across their epochs (field `trajectory`) |
 | `rank_ckpt_plan.json`, field `init_ckpt` | 5 | the BYOL checkpoints the five two-stage runs started from |
 | `superseded_ckpt_plan.json` | 30 | kept for the record; no number of the paper reads them |
 
@@ -41,9 +41,10 @@ script wrote it, with the rank score of every checkpoint of the run. The same fo
 
 Every tower that a number or a figure of the paper reads is here, at the checkpoint it is read
 at. A run saved a checkpoint every 50 epochs and the rank rule chose among them; the ones it did
-not choose are not here, except for the two runs listed under `trajectory`. The one other passage
-that follows runs over training, the view-count sweep at early epochs in Appendix C, quotes the
-readings the workers logged at each checkpoint.
+not choose are not here, except for the two runs listed under `trajectory`. The other passages
+that follow runs over training quote what the workers logged at each checkpoint: the learnable
+temperature and the view-count sweep in Appendix C, and the decline of the views-first anchors
+after epoch 50 in Supplement A.
 
 ## File names
 
@@ -76,16 +77,17 @@ row reads at least one of the row's runs. That file also names the notebook that
 | `i3ce_icetf_d512`, `i4ce_icetf_d512`, `ice_icetf_d512` | I-CE, λ = 3; I-CE, λ = 4; I-CE, λ = 1 | 512 | 0 to 2 | Table A4 |
 | `slot16i2cemean_icetf_d512`, `slot1i2cemean_icetf_d512`, `slot2i2cemean_icetf_d512`, `slot8i2cemean_icetf_d512` | I-CE, 16 latent queries; I-CE, 1 latent query; I-CE, 2 latent queries; I-CE, 8 latent queries | 512 | 0 to 2 | Table A6 |
 | `vfai2ce_icetf_d512` | I-CE, views-first anchors | 512 | 5 | Table S1 |
-| `ce_cetf_g1024`, `i2ce_icetf_g1024` | CE; I-CE, the recipe | 1,024 | 5 | Table A2, Table A3, Table A10, Table A14 |
-| `ce_cetf_g2048`, `i2ce_icetf_g2048` | CE; I-CE, the recipe | 2,048 | 5 | Table A2, Table A3, Table A10, Table A14 |
+| `ce_cetf_g1024`, `i2ce_icetf_g1024` | CE; I-CE, the recipe | 1,024 | 5 | Table A2, Table A10, Table A14 |
+| `ce_cetf_g2048`, `i2ce_icetf_g2048` | CE; I-CE, the recipe | 2,048 | 5 | Table A3, Table A10 |
 | `i2ce_icetl_g2048`, `i2cet05_icetf_g2048`, `i2cet10_icetf_g2048` | I-CE, learnable τ; I-CE, τ = 0.05; I-CE, τ = 0.10 | 2,048 | 5 | Table A3 |
 | `bce_cetf_gpool`, `byol_bytf_gpool`, `epd_v20i10c20_cetf_gpool`, `mq3072i2ce_icetf_gpool` | SimCLR-style; BYOL; VICReg (batch 192); I-CE with an EMA teacher and a memory bank | pool | 5 | Figure 7, Figure 8, Table 2, Section 3.2, Table A1, Table A12, Table A15, Table A16, Table A17, Figure A2, Table S2, Table S3, Table S4 |
-| `ce_cetf_gpool`, `i2ce_icetf_gpool` | CE; I-CE, the recipe | pool | 5 | Figure 7, Figure 8, Table 2, Section 6.1, Table A1, Table A2, Table A13, Table A14, Table A15, Table A16, Table A17, Figure A2, Table S2, Table S3, Table S4 |
+| `ce_cetf_gpool`, `i2ce_icetf_gpool` | CE; I-CE, the recipe | pool | 5 | Figure 7, Figure 8, Table 2, Section 6.1, Table A1, Table A2, Table A6, Table A12, Table A13, Table A14, Table A15, Table A16, Table A17, Figure A2, Table S2, Table S3, Table S4 |
 | `slot8i2cemean_icetf_gpool` | I-CE, 8 latent queries | pool | 5 | Table A6 |
 | `swin168step84loop2i2ce_icetf_gpool`, `swin168step84loop2i2ce_icetf_gpool_bw` | I-CE with the windowed teacher; I-CE, two-stage: BYOL warm start, then the windowed teacher | pool | 5 | Table A12, Table S2, Table S3, Table S4 |
 | `dino_dntf_gpool` | DINO at its published settings, the control of Section 4.3 | pool | 0 only | Table A1 |
 
-I + SimCLR (`i2bce_icetf`, files without a pack or a suffix, five folds) backs Figure 8. The
+I + SimCLR (`i2bce_icetf`, files without a pack or a suffix, five folds) backs Figure 8 and the
+five-fold readings of Appendix L. The
 fold-0 run of `i2ce_icetf_g2048` also gives the game vectors of Figure 10. The DINO-style control
 has one fold, so Table A1 gives its spread over the ten anchor draws; its file is 169 MiB because
 it carries the two prototype heads (see Format).
@@ -94,7 +96,7 @@ it carries the two prototype heads (see Format).
 
 | Runs | What they are | Pack | Folds | Read by |
 |---|---|---|---|---|
-| `ai25auni2_icetf_d512`, `ai2auni25_icetf_d512`, `ai4auni2_icetf_d512`, `ai6auni2_icetf_d512`, `i4uni2_icetf` | alignment + uniformity aimed at the anchor gallery, the four settings of Section 6.2 (`ai25auni2` is Table 3's a+u @ gallery row); alignment + uniformity aimed at the batch views (Table 3's a+u @ views row) | 512 | fixed split | Table 3, Section 6.2 |
+| `ai25auni2_icetf_d512`, `ai2auni25_icetf_d512`, `ai4auni2_icetf_d512`, `ai6auni2_icetf_d512`, `i4uni2_icetf` | alignment + uniformity aimed at the anchor gallery, the four settings of Section 6.2 (`ai25auni2` is Table 3's a+u @ gallery row); alignment + uniformity aimed at the batch views (Table 3's a+u @ views row) | 512 | fixed split | Table 3, Section 6.2, Appendix L |
 | `ai2ce_icetf_d512`, `ce_cetf_d512`, `i2ce_icetf_d512` | CE + anchor-in-I; CE; I-CE, the recipe | 512 | fixed split | Table 3, Section 6.2, Table A7 |
 | `ce_cetf_g2048`, `i2ce_icetf_g2048` | CE; I-CE, the recipe | 2,048 | fixed split | Figure 9, Figure A1, Figure A3, Appendix L |
 | `bce_cetf`, `i2bce_icetf`, `epd_v20i10c20_cetf` | SimCLR-style; I + SimCLR; VICReg, v/i/c = 20/10/20 (the recipe weights) | 512 | fixed split | Figure 9, Figure A3, Appendix L, Table A7 |
@@ -111,11 +113,14 @@ of the test curve Appendix L describes.
 Runs that read no anchor in training (`bce_cetf`, `i2bce_icetf`, `i4uni2_icetf` and the three
 VICReg weightings; the 512 in their rows is only the pack their worker encoded at each
 checkpoint) were selected after training: each of the run's 40 checkpoints encodes the
-4,096-sentence anchor pack of the fixed split and the rank rule chooses on that. These six runs
-and the fixed-split CE + anchor-in-I run stopped at 1,000 epochs under their first schedule and
-were trained on to 2,000 from the weights of epoch 1,000 with a fresh optimizer; the plan entries
-say so under `extended`. For CE + anchor-in-I the rule then selects epoch 1,650, and the
-epoch-1,000 file is kept and named under `previous`.
+4,096-sentence anchor pack of the fixed split and the rank rule chooses on that.
+
+Ten of the fifteen runs stopped at 1,000 epochs under their first schedule and were trained on to
+2,000 from the weights of epoch 1,000 with a fresh optimizer: those six, CE + anchor-in-I, and
+three of the gallery-aimed settings (`ai6auni2`, `ai4auni2`, `ai2auni25`). The plan entries say so
+under `extended`. Two selections moved with the longer runs, CE + anchor-in-I to epoch 1,650 and
+`ai4auni2` to epoch 1,250; the checkpoints selected before (epochs 1,000 and 900) are kept and
+named under `previous`.
 
 ## Format
 
@@ -177,7 +182,8 @@ repository give a run on the current packs that name themselves.
 Runs whose loss never reads the anchors (the SimCLR-style tower, I + SimCLR, BYOL, VICReg and
 alignment plus uniformity aimed at the batch views) train the same weights on either pack, so
 their checkpoints were not replaced; only readings that used the 512-sentence packs were redone.
-The fixed-split files of those runs therefore carry no `_d512`.
+Their files therefore carry no `_d512`: the fixed-split ones, and those of the five-fold
+I + SimCLR.
 
 Stripped hit@1 on the ten 4,096-sentence draws, five-fold mean ± std, before and after:
 
@@ -208,11 +214,13 @@ dataset keeps the first packs as `wscan_gal_rev_rawprefix.npz`; the workers read
 
 ## Changes
 
-- 2026-11-11, later that day. 61 checkpoints were added, so that every tower the paper reads is
+- 2026-11-11, later that day. 62 checkpoints were added, so that every tower the paper reads is
   here: the DINO-style control (fold 0), eight fixed-split runs at their selected epochs (CE and
   I-CE trained at 2,048 sentences, the SimCLR-style tower, I + SimCLR, a+u @ views and the three
-  VICReg weightings), and the 52 `trajectory` checkpoints. `rank_ckpt_plan.json` and
-  `fixed_split_ckpt_plan.json` name them; no earlier entry changed.
+  VICReg weightings), the 52 `trajectory` checkpoints, and epoch 1,250 of the 4:1 gallery-aimed
+  run of Section 6.2 (`ai4auni2_icetf_d512`). That run's entry in `fixed_split_ckpt_plan.json`
+  named epoch 900, its selection before it was trained on to 2,000 epochs; the paper reads epoch
+  1,250. No other earlier entry changed its checkpoint.
 - 2026-11-11. The four JSON files above replace the earlier per-campaign lists
   (`docfix512_ckpt_plan.json`, `c2_docfix512_ckpt_plan.json`, `c3_docfix512_ckpt_plan.json`,
   `q8_gpool_ckpt_plan.json`); no selection changed. The five BYOL checkpoints the two-stage runs

fixed_split_ckpt_plan.json

@@ -21,7 +21,8 @@
   "sha256": "778a94d36875894bc8872a43976a5f271152e8303605c1b4ca073ef9d6170ee3",
   "ranksel": 0.8912054361316923,
   "protocol": "fixed split (seed 0)",
-  "worker": "Pod/w9_a100_worker.py"
+  "worker": "Pod/w9_a100_worker.py",
+  "extended": "2026-11-08: trained on from 1,000 to 2,000 epochs (weights of ep1000, fresh optimizer, AMP and RNG); rank rule over all 40 checkpoints, which leaves the selection where it was"
  },
  "ai2ce_icetf_fixed_d512": {
   "arm": "ai2ce_icetf",
@@ -47,12 +48,19 @@
   "fold": null,
   "cap": "",
   "sfx": "_d512",
-  "ep": 900,
-  "ckpt": "ckpt_w9_wcle_ai4auni2_icetf_d512_fp_ep900.pt",
-  "sha256": "34f2ee0e051da230e57f713ec7446477686739d2eb4bac6e562ae606d0f876dd",
-  "ranksel": 1.5249424624108827,
+  "ep": 1250,
+  "ckpt": "ckpt_w9_wcle_ai4auni2_icetf_d512_fp_ep1250.pt",
+  "sha256": "c8dfd8a19a855f51d6a6c7afca635a2930d0312937c00538621f4d421bb05e64",
+  "ranksel": 1.6230743265482546,
   "protocol": "fixed split (seed 0)",
-  "worker": "Pod/w9_a100_worker.py"
+  "worker": "Pod/w9_a100_worker.py",
+  "extended": "2026-11-08: trained on from 1,000 to 2,000 epochs (weights of ep1000, fresh optimizer, AMP and RNG); rank rule over all 40 checkpoints",
+  "previous": {
+   "ep": 900,
+   "ckpt": "ckpt_w9_wcle_ai4auni2_icetf_d512_fp_ep900.pt",
+   "sha256": "34f2ee0e051da230e57f713ec7446477686739d2eb4bac6e562ae606d0f876dd",
+   "ranksel": 1.5249424624108827
+  }
  },
  "ai6auni2_icetf_fixed_d512": {
   "arm": "ai6auni2_icetf",
@@ -64,7 +72,8 @@
   "sha256": "d40740b355bd881d7865ca0ba004c99fff4f85b1147ec83ce090aa7ce5276996",
   "ranksel": 64.68700112813966,
   "protocol": "fixed split (seed 0)",
-  "worker": "Pod/w9_a100_worker.py"
+  "worker": "Pod/w9_a100_worker.py",
+  "extended": "2026-11-08: trained on from 1,000 to 2,000 epochs (weights of ep1000, fresh optimizer, AMP and RNG); rank rule over all 40 checkpoints, which leaves the selection where it was"
  },
  "ce_cetf_fixed_d512": {
   "arm": "ce_cetf",