发表机构
New York University; Columbia University; Zhejiang University（纽约大学; 哥伦比亚大学; 浙江大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出灰盒设置下的自适应自洽性方法，通过分布值观测的序列模式识别和ASC-D投注停止规则，在MMLU-Redux上显著减少推理轨迹并提高正确认证率。
AI 中文摘要
自洽性方法采样多条推理轨迹并聚合其最终答案，将大语言模型视为每条轨迹返回一个答案的黑盒。然而，每条轨迹的最终答案是从模型的log-probabilities中可获得的softmax向量中采样得到的。我们将此称为灰盒设置，其中每条轨迹揭示的是答案分布而非从中抽取的单个样本。我们在此设置中将高效推理表述为具有分布值观测的序列模式识别：一次采样一条轨迹，并在模型模态答案以指定置信水平被识别时立即停止。我们精确刻画了具有分布值观测的模式识别的渐近停止速率，并证明其绝不劣于黑盒速率。随后，我们提出了ASC-D算法，这是一种达到该渐近停止速率的投注停止规则。在MMLU-Redux上，ASC-D比仅答案的自适应自洽性基线少使用46.4%至95.6%的轨迹，并在三个开源模型上实现了最高的固定预算正确认证率。
英文摘要
Self-consistency samples many reasoning trajectories and aggregates their final answers, treating the LLM as a black box that returns one answer per trajectory. Yet the final answer of each trajectory is sampled from a softmax vector that is available from the model's log-probabilities. We refer to this as the grey-box setting in which each trajectory reveals this answer distribution rather than a single draw from it. We formulate efficient inference in this setting as sequential mode identification with distribution-valued observations: sample trajectories one at a time and stop as soon as the LLM's modal answer is identified at a prescribed confidence level. We characterize the asymptotic stopping rate of mode identification with distribution-valued observations exactly and show that it is never worse than the black-box rate. We then propose the ASC-D algorithm, a betting stopping rule that attains this asymptotic stopping rate. On MMLU-Redux, ASC-D uses $46.4$--$95.6\%$ fewer trajectories than answer-only adaptive self-consistency baselines and achieves the highest fixed-budget correct-certification rate across three open-source models.
Comments27 pages, 4 figures, 7 tables. The first two authors contributed equally
