arXiv cs.CL· Jiayi Li, Dennis L. Shung, Bradly C. Stadie·· 12 小时前AI 评分31
证据不足各有其因：让语言模型在不同证据状态下校准置信度
Each Insufficient in Its Own Way: Calibrating Language Models Across Varying Evidence States
AI 导读
研究者提出"证据状态校准"任务，要求模型在作答前的置信度与给定证据下答对的概率一致，并随证据变化而调整。据此提出的 SCE（充分性置信度估计）方法微调模型 next-token 分布中已有的置信度，在临床诊断和多跳问答上、两个模型族的四个模型中，Brier score 最多降低 60%，AUROC 最多提升 0.16。其置信度与经验成功率吻合，且能同时响应证据数量与具体证据内容。
正文
Abstract:How can a language model know whether it has enough evidence to answer? Language models tend to answer even when the evidence they have is insufficient. Deciding whether to answer or gather more evidence requires a calibrated estimate of how likely an answer is to be correct with the available evidence. However, most calibration work treats the input as fixed, without asking how this estimate should change with the evidence. We introduce \emph{evidence-state calibration}, which requires a model's confidence before answering to match its probability of answering correctly and to change accordingly whenever the evidence changes this probability. We propose \ours{} (\textbf{Suffi}ciency \textbf{C}onfidence \textbf{E}stimation), which fine-tunes the confidence already present in a model's next-token distribution to match how often the model answers correctly from each evidence state. Across clinical diagnosis and multi-hop question answering with four models from two families, \ours{} reduces the Brier score by up to 60\% and increases AUROC by up to 0.16. Its confidence matches the empirical success rate and responds not only to the amount of evidence but also to which evidence is present, showing that models can learn to judge whether their current evidence is enough to answer.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.33237 [cs.CL] |
| (or arXiv:2610.33237v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.33237 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jiayi Li [view email]
[v1]
Fri, 23 Oct 2026 19:39:08 UTC (637 KB)
来源：arXiv cs.CL · arxiv.org