research: P2 三臂对照（B1 主实验）+ E5 对比探针负结果 - #336
Open
oratis wants to merge 1 commit into
Open
Conversation
三臂对照(P1 程序化 40 items,冻结 Qwen2.5-1.5B/3B,零训练): - 压缩臂(我方门):核心域 23/24 AUC(z) 0.911(1.5B)/22/24 0.885(3B) ——唯一能读出消歧证据的判据 - 语义熵臂(候选条件化 P2b):AUC 0.271/0.212(核心域),且 SE(M′)≥SE(M) ——比 DESIGN §2 预测更强:SE 不仅分不开,还系统性偏爱更具体的误解 (3B 核心域两组一致性双双饱和 1.000) - 后验/置信臂(EIG 类):教学后 forced-choice 判对仅 4/40(1.5B)/11/40(3B), 自信错误 14-17/40——教学集构造不区分时后验由 size-principle 先验决定, 「按置信度接受」会大规模固化误解。C4 独赢点的直接实验形态 P0k 负结果(E5 对比条件化解法):把全部消歧证据放进 prompt 做显式二选一 仍押 M′(1.5B 核心域 0/24)——同样证据,概率层测量(A 臂)提取得出, 言语化判断提取不出 ⟹ 门必须读概率、不能问模型。归档(同 F(c) 探针) 诚实 caveats(p2/README):P1-G5 材料未复现 p0-G5 的 E5 失败(偏易), E5 测试床仍是 p0 材料;自由行为版 B 臂有混淆,论文只引 B′; 单模型族、每类 n=8、无多重比较校正 文档:REPORT §3.4b 新增;DESIGN_COMPRESSION_GATE §3#10 归档对比条件化 失败、§4 B1 行标注已跑 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This was referenced Aug 1, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
内容
P2 三臂对照（DESIGN_COMPRESSION_GATE §4 的 B1 主实验）——把三个候选准入判据摆在同一材料（P1 程序化 40 items）上，冻结 Qwen2.5-1.5B/3B，零训练：
P0k 负结果：E5 的「对比条件化」解法失败——证据全给、显式二选一仍押 M′（核心域 0/24）。同样的证据，概率层测量提取得出，言语化判断提取不出 ⟹ 门必须读概率。已归档避免重复（同 P0i 忠实性探针）。
诚实 caveats（详见
p2/README.md）：P1-G5 未复现 p0-G5 的 E5 失败（材料偏易），E5 测试床仍是 p0；单模型族、n=8/类。🤖 Generated with Claude Code