推理模型中轨迹长度是一个简单的不确定性信号
Trace Length is a Simple Uncertainty Signal in Reasoning Models
AI总结：
本研究表明推理轨迹长度是大型推理模型中简单且有用的置信度估计器，其效果与言语化置信度等零样本方法相当且互补，并揭示了推理后训练从根本上改变了轨迹长度与准确率的关系，确立了轨迹长度作为实用置信度度量的地位。
摘要：
LLM 的不确定性量化是解决幻觉及其他限制其可靠部署问题的关键研究方向。在本研究中，我们表明推理轨迹长度是大型推理模型中一种简单且有用的置信度估计器。通过跨多个模型、数据集和提示词的综合实验，我们表明轨迹长度以与其他零样本置信度估计器（如言语化置信度）相当但互补的方式发挥作用。我们的工作揭示了推理后训练从根本上改变了轨迹长度与准确率之间的关系，超越了先前表明后训练会导致轨迹总体变长（例如“过度思考”）的研究。我们调查了轨迹长度作为置信度信号表现背后的机制，观察到即使在调整了问题难度和 GRPO 引起的长度偏差等混杂因素后，该效应仍然存在。我们识别出高熵或“分叉”token 在该机制中起着关键作用。我们的发现表明，推理后训练增强了超越言语表达的不确定性量化，并将轨迹长度确立为大型推理模型的实用置信度度量。
英文摘要：
Uncertainty quantification for LLMs is a key research direction towards addressing hallucination and other issues that limit their reliable deployment. In this work, we show that reasoning trace length is a simple and useful confidence estimator in large reasoning models. Through comprehensive experiments across multiple models, datasets, and prompts, we show that trace length performs in comparable but complementary ways to other zero-shot confidence estimators such as verbalized confidence. Our work reveals that reasoning post-training fundamentally alters the relationship between trace length and accuracy, going beyond prior work that had shown that post-training causes traces to grow longer in general (e.g., "overthinking"). We investigate the mechanisms behind trace length's performance as a confidence signal, observing that the effect remains even after adjusting for confounders such as problem difficulty and GRPO-induced length bias. We identify high-entropy or "forking" tokens as playing a key role in the mechanism. Our findings demonstrate that reasoning post-training enhances uncertainty quantification beyond verbal expressions, and establish trace length as a practical confidence measure for large reasoning models.
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。
