REHEARSE：大语言模型中用于语言置信度校准的经验式排练
REHEARSE: Experiential Rehearsal for Verbal Confidence Calibration in Large Language Models
AI总结：
针对大语言模型置信度与实际正确性不匹配的问题，提出无训练的Rehearse方法，通过置信度校准博弈的反馈生成校准信号，在多模型多基准实验中显著降低了预期校准误差并提升准确率。
摘要：
大语言模型（LLMs）常出现语言置信度与实际正确性不匹配的情况，这限制了它们在安全关键型应用中的可靠性。现有的基于提示的方法大多将校准视为一次性推理问题，依赖实例级推理或事后自我评估。我们提出Rehearse（Experiential Rehearsal，经验式排练），这是一种无训练的方法，可让模型从自身带评分的置信度经验中进行自适应。在基于严格适当评分规则的置信度校准博弈中，模型会收到先前置信度决策的反馈；该经验被总结为博弈后轨迹前缀，用于捕捉系统性的过度置信或置信不足。推理时，模型将这种跨实例校准信号应用于每个新问题的思维链推理轨迹。在四个LLMs、三个基准和五个随机种子的实验中，Rehearse在无训练方法中实现了最低的平均预期校准误差（ECE），同时准确率有所提升，相对于未校准基线将平均ECE降低了58%。代码可在该https URL获取。
英文摘要：
Large language models (LLMs) often express verbal confidence that is poorly aligned with actual correctness, limiting their reliability in safety-critical applications. Existing prompt-based methods treat calibration largely as a one-shot inference problem, relying on either instance-level reasoning or post-hoc self-assessment. We introduce Rehearse (Experiential Rehearsal), a training-free method that instead enables models to adapt from their own scored confidence experience. In a credence-calibration game grounded in a strictly proper scoring rule, the model receives feedback on prior confidence decisions; this experience is summarized in a post-game trajectory prefix that captures systematic over- or under-confidence. At inference time, the model applies this cross-instance calibration signal to the chain-of-thought reasoning trace for each new question. Across four LLMs, three benchmarks, and five random seeds, Rehearse achieves the lowest average ECE among training-free methods with improved accuracy, reducing average ECE by 58% relative to the uncalibrated baseline. Code is available at https://anonymous.4open.science/r/Experiential-Rehearsal-7C77/.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
