evil_MoE

mirror of https://github.com/wassname/evil_MoE.git synced 2026-06-27 17:48:43 +08:00

Files

T

wassname ea01267cd8 fix: eval on paper test set, not contaminated holdout (base solve 0.94->0.094)

The periodic VAL eval ran on leetcode_train_medhard_holdout.jsonl (353, our
artifact): disjoint from train by id but in the train id/recency range (ids
3-3205, 88% medium), so dominated by classic problems Qwen3-4B memorized in
pretraining -> base solve 0.94, saturating solve and killing the hack metric's
gt-fail headroom. Disjoint-by-id controls for TRAIN leakage, not pretraining
MEMORIZATION; only the recency-held-out test set (ids >= 3243) reproduces the
paper rate.

Proof (job 176, base model, same eval_hack_solve): test_medhard solve=0.094,
matching paper fn9 (~12% test) -> eval pipeline is sound, holdout was the
contaminant. Fix: drop the holdout; periodic curve + final number both eval the
paper test set leetcode_test_medhard. Smoke green. Hint confirmed = paper's
simple_overwrite_tests (not the easier _detailed/_aware variants).

Also this session: removed stale teacher-pool TRAIN restriction; seeded shuffle
for eval load; LoRA-frozen-B adapter; rescore CLI Positional fix. Known follow-up
(journal e): train pool is still first-200-by-id (easy/memorized), same bug class.

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

2026-06-07 11:01:31 +00:00

attic

retract 'null_city contaminated' framing -> in/out-of-subspace + cosine-is-correlational

2026-06-05 09:21:41 +00:00

build_combined_pool.py

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

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

build_substrate.py

rename python package projected_grpo -> vgrout

2026-06-05 14:51:48 +08:00

make_random_vhack.py

cleanup: trim 2 stale provenance/train-of-thought comments