arXiv cs.CL· Chu Fei Luo, Samuel Dahan, Xiaodan Zhu·· 9 小时前AI 评分31
把提问当作隐藏状态探针:LLM 测试时推理中的自我诊断与纠错鸿沟
What Am I Missing? Question-Answering as Hidden State Probing
AI 导读
研究将“向教师模型提问”作为推理时干预,在学生模型提问前后的隐藏状态上训练探针,发现该探针在教师给出答案前就能预测最终正确性,说明提问过程本身带有自我诊断信号。作者进而把提问建模为序列决策问题,用探针作质量分数定义门控策略以最大化正确率。实验显示提问干预的效果高度依赖模型自我一致性,且诊断与纠错之间存在鸿沟:干预伤害正确轨迹的概率与挽回错误轨迹相当。
正文 · 原文
Abstract:The study of test-time reasoning in large language models (LLMs) has attracted increasing interests due to its profound impact on many tasks, with early work tracing back to approaches such as chain-of-thought reasoning. The basic mechanism underlying the test-time reasoning process, however, is still under-studied: from the same input prompt, and even the same partial solution, LLMs can produce varied answers if sampled multiple times. In this paper, we leverage question-asking as an inference-time intervention that articulates information about the model's hidden state. Specifically, we present a student-teacher setting where a student asks questions to a teacher. We train a probe on the student's hidden state before and after asking a question and find it is predictive of the trajectory's final correctness, even before generating the teacher's answer. This suggests there is a meaningful signal from the self-diagnosis that occurs during question generation rather than information transfer from the teacher. We then frame question-asking as a sequential decision problem, using this probe as a quality score, and define a gating policy to ask questions that maximize likelihood of correctness. We find that the success of question-asking as an intervention is largely dependent on the model's self-consistency. Our empirical results show a gap between detection and recovery; while our gating policy captures model correctness and uncertainty, interventions are equally likely to harm correct trajectories as they are to recover incorrect ones. This gap between diagnosis and correction has broader implications on language models' capacity for self-refinement under uncertainty.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2605.31561 [cs.CL] |
| (or arXiv:2605.31561v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.31561 arXiv-issued DOI via DataCite |
Submission history
From: Chu Fei Luo [view email]
[v1]
Fri, 29 May 2026 17:27:07 UTC (200 KB)
[v2]
Mon, 16 Nov 2026 19:48:28 UTC (256 KB)
来源:arXiv cs.CL · arxiv.org