arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.20508cs.MAcs.AIcs.CL
收藏

衡量推理模型的弱到强可读性

Measuring Weak-to-Strong Legibility of Reasoning Models

University of California, Berkeley(加州大学伯克利分校);Stanford University(斯坦福大学)

Dani Roytburg, Shreya Sridhar, Daphne Ippolito

AI总结:

针对推理语言模型在多智能体场景中生成的中间思维链,提出“弱到强可读性”概念,并设计衡量指标以评估强模型输出对弱模型的易理解性。

摘要:

推理语言模型及其生成的中间思维链在多智能体设置(如模型间监控或蒸馏到较小模型)中扮演着越来越核心的角色。当不同能力层级的智能体必须合作时,强模型需要产生能被弱模型消化的轨迹。我们将此目标称为“弱到强可读性”。大模型的可信度部分依赖于这种可读性属性。特别是在安全监督方面,采用弱监控器可能成为健康预算下可靠性支架的标准。可读性要求这些决策轨迹的形状采取某种弱监控器可访问的形式。现有的基于效率的可读性指标未能捕捉“彻底性”,而是侧重于简洁性。

英文摘要:

Language models are increasingly trained to "reason" before answering users' queries, outputting hundreds of intermediate tokens before a final answer. While reasoning is designed with direct model capabilities in mind, we argue that reasoning language models (RLMs) should also be assessed by interaction effects between their reasoning traces and weaker models. In contexts like safety monitoring and distillation, the performance of weak models increasingly depends on legibility--how thoroughly and precisely a reasoning trace externalizes an RLM's actions. Existing efficiency-based metrics for legibility fail to capture "thoroughness", instead focusing on conciseness. Thus, we introduce transfer utility, a method for measuring trace legibility derived from interactions where weak models finish tasks given incomplete traces. Evaluating 84k traces from 12 RLMs across three tasks, we find that reasoning traces generated by the highest-performing and the most efficient models rank the lowest for transfer utility. We also find evidence that transfer utility predicts the ability of weak monitors to verify procedurally dense traces for math or logical problem-solving, though the effect moderates when verifying fact-driven traces. Finally, we show that open-source reward models decouple from transfer utility signals on correctness. Together, these findings surface status quo trade-offs in weak-to-strong legibility, motivating an interaction-first framework for future applications.

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。

↑