arXiv cs.AI· Khushal Sethi·· 4 小时前AI 评分31
一致性不等于纠错:拆解 LLM 智能体的置信度、弃答与恢复
Consistency Is Not Correction: Separating Confidence, Abstention, and Recovery in LLM Agents
AI 导读
研究将 LLM 智能体的置信度信号拆为排序、选择性预测与纠错效用三个互不蕴含的目标,并给出纠错效用的精确分解:门控恢复减去门控伤害。在 840 回合受控实验中,没有任何验证器在 60 个 GSM8K 决策上替换原提案,使该对比实际未检验任何机制;小模型套件中扰动一致性验证器达到策略内 AUROC 0.990,但每次结果与单样本解码一致,调用次数为 4.32 倍。
正文 · 原文
Abstract:A controller that resamples an answer, escalates to a larger model, or abstains is usually justified by a confidence signal. Three claims about such a signal do not follow from one another: it ranks answers (discrimination), it supports withholding some of them (selective prediction), and it changes an answer for the better (correction utility). We give an exact decomposition of the third into gated recoveries minus gated harms, which bounds whether a comparison can succeed before it is run: a gate that only replaces a proposal can act only where the base policy was already wrong, so the initial error count caps the discordant pairs any correct gate can produce. When that cap falls below the count a paired test needs, the experiment cannot return a positive result for a working mechanism, nor distinguish one from a gate that never fired. We measure all three targets on two model scales and find the bound binds. In a small-model suite a perturbation-agreement verifier reaches within-policy AUROC 0.990 (95% interval [0.95, 1.00]) while every episode outcome matches single-sample decoding at 4.32 times the call count. In a controlled 840-episode study no verifier replaced the proposal on any of the 60 GSM8K decisions, so that comparison tested nothing; the entire GSM8K gain across six controllers is one task, and it is the task whose identical temperature-zero repeats disagreed, making it the one answer the service does not reproduce. On the diagnostic set the gate could reach four errors while the test needed six, so even a perfect gate had no significant result available. We report two checks that would have caught this before any budget was spent, and show that tie handling in rankings, tied selection cutoffs, and unequal output limits each move a headline number in the same dataset. The binding constraint on this literature is measurement resolution, not the choice of signal.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2611.21651 [cs.CL] |
| (or arXiv:2611.21651v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.21651 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Khushal Sethi [view email]
[v1]
Thu, 15 Oct 2026 05:55:30 UTC (61 KB)
来源:arXiv cs.AI · arxiv.org