diff options
| author | Yuren Hao <yurenh2@illinois.edu> | 2026-08-04 04:13:43 -0500 |
|---|---|---|
| committer | Yuren Hao <yurenh2@illinois.edu> | 2026-08-04 04:13:43 -0500 |
| commit | 3962588a667d0d8945044b105dedda75e3fa3a41 (patch) | |
| tree | 923698a7dacada2770c2c94860a3243b1496a04e /ep_run/train.py | |
| parent | 1d2d0338ae25de7d60756545dad692c2f833c532 (diff) | |
RESULT 89: 135M判决 — 修复版3.2024 vs BP双seed均值3.2054(gap -0.0030=seed间距一半), 统计不可分
旧读出同点+0.3703=+44.8%ppl, 即"宽度墙"真实体量, 全部由一行读出改写消除。五尺寸(11M-135M)
修复后全部落入种子噪声。BP s2完赛best 3.0968。EP剩11.4k步出终局。
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
Diffstat (limited to 'ep_run/train.py')
0 files changed, 0 insertions, 0 deletions