wassnameandClaudypoo f0cbbacaf0 save per-eval deploy-adapter ckpts (rescore w/o retrain) + CLAUDE.md test lesson ...

save_eval_ckpts (default on): write the deploy adapter (δS only, ~2.3MB) at each
deploy-eval step, step-tagged, so a run can be re-scored later (more prompts /
different eval) without retraining. The A5 run saved only final+first_hack, which
is why the leak needed a full retrain rather than a rescore.

AGENTS.md: every load-bearing invariant gets a verify_*.py gate. The no-cheat leak
shipped because the green gates never covered the property -- 'tests passed' is
meaningless if the property was never tested.

Co-Authored-By: Claudypoo <288921227+claudypoo@users.noreply.github.com>

2026-06-05 03:58:26 +00:00

..

__init__.py

setup

2026-05-23 10:40:02 +08:00

antipasto.py

refactor: decompose train.py helpers into clean's module names

2026-06-01 12:15:12 +00:00

data.py

refactor(c): extract data.py (load_problems) + vhack.py (load/postprocess v_hack)

2026-06-01 09:29:03 +00:00

derisk_loopholes.py

refactor: decompose train.py helpers into clean's module names

2026-06-01 12:15:12 +00:00

eval.py

cleanup: trim stale comments + attic README

2026-06-03 00:09:19 +00:00

extract_vhack_grad.py

style: em-dash -> ASCII '--' in comments across src (check-1 dir-wide)

2026-06-01 12:20:54 +00:00

figs.py

plots: dejargon, drop redundant titles, emit png+svg+pdf, CSV re-render

2026-06-03 02:44:39 +00:00

pairs_from_pool.py

reorg: out/ sorted by datatype (vhack/ pools/ runs/ vhack_grads/ figs/)

2026-05-30 03:52:24 +00:00

pairs.py

style: em-dash -> ASCII '--' in comments across src (check-1 dir-wide)

2026-06-01 12:20:54 +00:00

problems.py

refactor: decompose train.py helpers into clean's module names

2026-06-01 12:15:12 +00:00

proj.py

cleanup: trim stale comments + attic README

2026-06-03 00:09:19 +00:00

regrade_pool.py

style: em-dash -> ASCII '--' in comments across src (check-1 dir-wide)

2026-06-01 12:20:54 +00:00

rewards.py

style: em-dash -> ASCII '--' in comments across src (check-1 dir-wide)

2026-06-01 12:20:54 +00:00

tablelog.py

perf(eval): drop redundant per-step knob-ON pass, default eval-every 5->10

2026-06-04 02:25:07 +00:00

train.py

save per-eval deploy-adapter ckpts (rescore w/o retrain) + CLAUDE.md test lesson

2026-06-05 03:58:26 +00:00

vhack.py

refactor(c): extract data.py (load_problems) + vhack.py (load/postprocess v_hack)

2026-06-01 09:29:03 +00:00