Everything behind Sections 2, 3, 6, 7 and 9 of the CoSER reproduction report: the check scripts and the platform source as run, their results and logs, the outputs of the local end-to-end runs, the anchor and the judge recheck, and the three independent reviews. The checks call no model; the anchor and the recheck, which did, are copied as they were written. Nothing here contains test-set text: the data stay with their authors (github.com/Neph0s/CoSER); lines of the reviews that quoted them are replaced by a marker (logs/06-redaction.log). The reviews are by AI agents the platform's operator ran; no outside party has reviewed this. The platform's CoSER leaderboard was withdrawn on 2026-11-01; the report says why.
sha256sum -c SHA256SUMS
| File | Bytes | SHA-256 | What it is |
|---|---|---|---|
| code/docs/复现规范/CoSER-核查矩阵.md | 35,961 | 2acc4ec602d8… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_anchor.py | 7,238 | 2231bcca8d86… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_audit_bundle.py | 17,284 | 557162d1fa4e… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_e2e_check.py | 13,329 | 5ae64ac75a5b… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_e2e_mock.py | 7,440 | 97cf15509ac3… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_gca_dump.ts | 9,103 | 665d8212b29e… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_judge_recheck.py | 11,395 | 2956cd40b439… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_method_validity.py | 10,397 | 91cd886f6591… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_paper.py | 3,328 | b89ffb8bf909… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_prompt_check.py | 10,797 | f26811be0eb6… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_reference.py | 69,312 | 695d951c508d… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_rouge_check.py | 4,167 | 77913d275704… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_spec_check.py | 54,394 | 2f5be9b0c289… | The check scripts, the matrix and the platform source they execute, as run. |
| code/tools/experiments/coser_table_checks.py | 16,062 | 434a8a1a398c… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/data.ts | 2,566 | 54c15e1645da… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/aggregate.ts | 2,316 | 2cb9f56444d2… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/bleu.ts | 13,445 | 00dab41e8741… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/crmath.ts | 3,539 | 02c12ffd00a8… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/family.ts | 3,447 | 5c3427959b25… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/index.ts | 1,773 | e13c5ecd2400… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/judge.ts | 9,602 | 85d032930185… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/loop.ts | 6,895 | 5f2df0c2400a… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/punkt.gen.ts | 280,341 | 18aef32a032e… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/py.ts | 13,200 | 0a52810a3b18… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/rng.ts | 1,993 | 23218f35f334… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/setup.ts | 8,648 | 66867e0e7b3e… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/strings.gen.ts | 16,806 | 08f66e9d3bf2… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/gca/text.ts | 6,805 | 64581ef594a4… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/index.ts | 16,925 | 5833e22e07f2… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/meta.ts | 2,680 | 16eabd9d1186… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/benchmarks/coser/published.ts | 8,102 | 9cd914f07657… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/engine/results.ts | 31,701 | 995baf8e125d… | The check scripts, the matrix and the platform source they execute, as run. |
| code/worker/src/scoring.ts | 4,095 | 41666b1b07e8… | The check scripts, the matrix and the platform source they execute, as run. |
| logs/00-environment.log | 639 | 226d0f26c068… | One log per step: command, start and end time (UTC), exit status, full output. |
| logs/01-spec_check.log | 4,547 | 66f51dee873f… | One log per step: command, start and end time (UTC), exit status, full output. |
| logs/02-prompt_check.log | 2,396 | 7593c4a0acb5… | One log per step: command, start and end time (UTC), exit status, full output. |
| logs/03-paper.log | 551 | eaffcab0e2b7… | One log per step: command, start and end time (UTC), exit status, full output. |
| logs/04-table_checks.log | 3,914 | 4a03352cb845… | One log per step: command, start and end time (UTC), exit status, full output. |
| logs/05-rouge_check.log | 1,333 | f458b79c815d… | One log per step: command, start and end time (UTC), exit status, full output. |
| logs/06-redaction.log | 1,792 | eca60a49aee3… | One log per step: command, start and end time (UTC), exit status, full output. |
| results/coser_anchor.result.json | 5,929 | 5e3cee254103… | The anchor (V09, not reproduced): verdict, scores, splits, run statistics, lengths, line counts, cost. |
| results/coser_anchor.rows.json | 254,981 | 19196b24a471… | The anchor by session: scene id, run, lines, scores, severity sums, flaw counts; numbers only. |
| results/coser_judge_recheck.result.json | 5,311 | eaf613c9d481… | The authors' three released simulations judged today against their stored critiques (V09a). |
| results/coser_judge_recheck_variants.result.json | 4,352 | 9d485a653321… | Character fidelity on the same simulations under variants of the judge (V09a). |
| results/coser_method_validity.result.json | 2,270 | d415c36f2280… | What the score does on the anchor's sessions: repeatability, the length term, the floor, against the paper's differences between models (Section 7.4). |
| results/coser_original_as_simulation.result.json | 11,458 | 238484a59e79… | The book's own conversation judged as a simulation, against GPT-4o on the same scenes (V09b; numbers only). |
| results/e2e/out_extra.txt | 146 | 22e86ffe1866… | Temperature, limit, top-p and seed placed in the settings: what the model under test received (V14). |
| results/e2e/out_faults.txt | 274 | 8c9194eb6a6f… | Failure injection: empty replies, refusals, a transient 500; temperature sequences and stored lines (V06). |
| results/e2e/out_full.txt | 66,170 | 4629438fd9f2… | The full configuration: 600 sessions with stand-in models; eligibility under the anchor flag. |
| results/e2e/out_gateway403.txt | 116 | 73f4f41e3e83… | A gateway 403 whose text matches the moderation pattern: the session fails, no scored line (V06). |
| results/e2e/out_regular.txt | 2,711 | 2141588eeaa8… | 18 sessions of a local run: each replayed through the reference; requests, temperatures, limits, records, scores and aggregates. |
| results/prompt_check.json | 5,339 | 72f07d5fe8fd… | The paper's prompts against the platform's strings, with every difference and its meaning. |
| results/rouge_check.json | 1,574 | bdf044ef838b… | ROUGE-L: every release of the rouge package on the released values. |
| results/spec_check.txt | 4,376 | 15c2da4d5a12… | The protocol module against the official code. |
| results/table_checks.json | 22,665 | 994fb56ed6d4… | The paper's own numbers and the released files. |
| reviews/coser-pass1-paper-first.md | 46,966 | 87a532448947… | First independent review: from the paper to the platform. |
| reviews/coser-pass2-code-first.md | 51,171 | d0376b806882… | Second independent review: from the official code to the platform. |
| reviews/coser-pass3-numbers.md | 53,192 | d3036cda29d4… | Third independent review: every number of the report recomputed from its sources. |
| reviews/coser_third_party_search.md | 20,457 | 4c781179c514… | The search for later work reporting CoSER scores (Section 9): queries, sources and what was verified. |
| reviews/pass1-scripts/keyword_scan.py | 1,565 | 16ef4587f5a1… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/keyword_scan.txt | 37,559 | b11575cbe6dd… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/platform_on_example.ts | 2,782 | 4ce8f9827794… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/platform_on_example.txt | 3,618 | 6864ad872dc5… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/prompt_compare.py | 5,264 | 6b7585e94e79… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/prompt_compare.txt | 5,027 | 9cfd18b9e299… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/published_eval.py | 6,134 | 93a7221bfd2e… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/published_eval.txt | 2,340 | 447d52eb1b73… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/strings_vs_official.py | 3,069 | 4a544d9d9eab… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/strings_vs_official.txt | 1,413 | 23687c46a66b… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/table_check.py | 6,571 | 5701b0b3aed1… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/table_check.txt | 5,220 | fccc816b0193… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/testset_check.py | 3,820 | f44866daf415… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/testset_check.txt | 1,462 | f95db0c413a3… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/version_diff.py | 1,625 | d74f936253fb… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass1-scripts/version_diff.txt | 31,895 | f09e9b9401e7… | The scripts and outputs of review 1, without its copies of the authors' files. |
| reviews/pass2-scripts/01_data.py | 4,364 | 8966786316fe… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/02_exp.py | 5,320 | 3ed5c09ab0b9… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/03_replay.mts | 2,110 | c10d39e49db7… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/04_strings.py | 6,334 | 331570f996ec… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/04a_dump.mts | 293 | a48c60341137… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/05_versions.py | 1,524 | 3e52c5a62deb… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/06_cmp.mts | 1,568 | f4d29a650012… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/06_ref.py | 8,974 | a6d779e89b15… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/07_py_sim.py | 10,592 | 62fb49ee9209… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/07a_ts_sim.mts | 2,612 | 314f66f5c6d8… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/08_estimate.mts | 1,495 | 0f015c0494aa… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/09_client_mock.mts | 3,688 | f235b0c87a27… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/10_paper.py | 2,891 | 7986de65342f… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass2-scripts/10a_pub.mts | 170 | fcf139e7529c… | The scripts and outputs of review 2, without its copies of the authors' files. |
| reviews/pass3-scripts/cost.py | 1,457 | 9c55ffd97d54… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/data_counts.py | 6,308 | 4d0bab745405… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/e2e_logs.py | 1,755 | be9d52e00a75… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/heldout.py | 1,365 | aad9d54a85c0… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/heldout2.py | 925 | c87621aa187c… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/htmltext.py | 381 | cf489b6bfd18… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/issues.txt | 19,804 | 6a41881dd5f2… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/nums.py | 2,471 | ba817f927bdc… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/out/data_counts.txt | 1,927 | 9da51deb6c25… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/out/prompt_check.txt | 2,198 | af995e2e1fc7… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/out/spec_check.txt | 4,383 | 0783d888ac0f… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/out/table_checks.txt | 3,702 | d9c5e596d89d… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/sha_prompts.py | 1,126 | f3aca2c79eaf… | The scripts and outputs of review 3, without its copies of the authors' files. |
| reviews/pass3-scripts/table3_count.py | 378 | e401f853c116… | The scripts and outputs of review 3, without its copies of the authors' files. |