通过外化不确定性，让大模型在推理中自我修正。
Understanding Reasoning in LLMs through Strategic Information Allocation under Uncertainty
- 将推理拆分为进展与不确定性表达，以信息论框架建模
- 仅用少量怀疑提示即可恢复失败推理路径，成功率显著提升
- 适合研究模型反思机制或改进推理能力的研究者
大模型常出现如‘等等’后自我修正的顿悟时刻，但其机制尚不明确。标准大模型主要因隐性偏离而崩溃：轨迹虽偏离正确答案，但仍保持局部连贯性，无显式错误信号触发自我修正。本文提出一种信息论框架，将推理分解为过程推进与认知外化（即在词元层面表达不确定性），并证明零星的不确定性表达可使轨迹重新收敛至正确答案，即使无显式错误提示。实验表明，极简的怀疑提示即可恢复失败轨迹，小规模监督微调足以激活或抑制该能力，说明强推理并非依赖超凡内核机制，而更多取决于语言习惯中的不确定性外化。本框架将推理重新定义为不确定条件下的策略性信息分配，为理解与提升大模型推理提供了新视角。
原文摘要 · Abstract (English)
LLMs often exhibit Aha moments such as self-correction after tokens like "Wait," yet the underlying mechanism remains unclear. Standard LLMs collapse mainly through silent divergence, where trajectories drift from the correct answer yet remain locally coherent, so no explicit error triggers reactive self-correction. We introduce an information-theoretic framework that separates reasoning into procedural advancement and epistemic verbalization, the token-level externalization of uncertainty, and prove that sporadic verbalization restores convergence toward the correct answer even without explicit error triggers. Empirically, a minimal doubt cue recovers failed trajectories, and small-scale SFT suffices to instill or suppress this capability, suggesting that strong reasoning hinges less on an extraordinary inner mechanism than on the linguistic habit of externalizing uncertainty. Our framework recasts reasoning as strategic information allocation under uncertainty, offering a new lens for understanding and advancing LLM reasoning.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。