diff options
| author | Yuren Hao <yurenh2@illinois.edu> | 2026-08-02 13:44:50 -0500 |
|---|---|---|
| committer | Yuren Hao <yurenh2@illinois.edu> | 2026-08-02 13:44:50 -0500 |
| commit | 42c9dddb6a0c176f617b53233a1bff8f91e1c7cb (patch) | |
| tree | f3d5ee20fe81c8df91e6ecbbc2611098bf6ccf12 /ep_run/eig_traj2.py | |
| parent | 5d906affa10d657d0d390b1ae472474bff5b9068 (diff) | |
审计: EP/coupled-learning公开实现的对比读出地板普查(33仓库)
失效律 rel_err ≈ 0.3-3×eps32/r 已定量(r=1e-5→1-3%, 1e-6→30%, 1e-7→噪声)。
结论: 模式近乎普遍(13/20命中含Scellier原版/Laborieux/Ernoult/Rain)但领域"侥幸安全"
(物理侧fp64, ML侧β≥0.1); 无一仓库对r设断言或用fp64累加对比量; JAX/GPU迁移正在侵蚀余量。
Tier1唯一危险者=Rasmuskh/dualprop_icml_2024(VGG16×β=0.01×fp32,输出层可1行修复),
已本人clone亲验并修正子代理表述(隐藏层经act非逐字替换)。
修正我方假设: 全息EP非结构免疫(消截断偏差不消舍入)。
诚实负面: 深度驱动r塌陷仅卷积成立, MLP下r反增, 不得当普遍律。
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
Diffstat (limited to 'ep_run/eig_traj2.py')
0 files changed, 0 insertions, 0 deletions