ChainUQ：面向大语言模型的推理一致性感知不确定性量化
ChainUQ: Reasoning Consistency-Aware Uncertainty Quantification for Large Language Models
查看机构详情
- Florida State University（佛罗里达州立大学）
机构由 AI 辅助整理，请以论文原文为准。
浏览论文内容
中文总结 AI 辅助
ChainUQ提出推理一致性感知的不确定性量化框架，通过对齐感知轻量级模块和一致性校准器，提升LLM响应级置信度可靠性，在AUROC和ECE上显著改善且可迁移。
中文摘要 AI 辅助
尽管大语言模型（LLMs）展现出令人印象深刻的推理能力，但当中间论断与最终结论相冲突时，其响应级别的置信度可能仍然不可靠。因此，需要有效的不确定性量化（UQ）来捕捉推理链中的逻辑不一致性，而不仅仅是最终输出的正确性。现有方法存在两个主要局限：（1）它们依赖词元级别的概率，无法捕捉推理一致性；（2）它们缺乏利用生成链的结构逻辑动态校准置信度的机制。为推进现有研究，我们提出了ChainUQ，一个面向大语言模型的推理一致性感知不确定性量化框架。ChainUQ包含两个关键技术组件：一个对齐感知的轻量级UQ模块，它从与最终结论对齐的冻结特征中估计原始的内在模型置信度分数；以及一个推理一致性感知校准器，它利用推理链一致性证据来细化该分数。在多种分布内和分布外基准上的评估表明，ChainUQ持续改善了响应级别的不确定性估计，在AUROC上平均相对提升3.1%，在ECE上相对降低高达45.0%，并且无需额外微调即可直接迁移到新设置中。
英文摘要
While large language models (LLMs) exhibit impressive reasoning capabilities, response-level confidence may remain unreliable when intermediate claims conflict with the final conclusion. Therefore, effective uncertainty quantification (UQ) is required to capture logical inconsistencies within the reasoning chain, not just the correctness of the final output. Current approaches have two major limitations: (1) their reliance on token-level probabilities fails to capture reasoning consistency, and (2) they lack mechanisms to dynamically calibrate confidence using the structural logic of the generated chain. To advance existing research, we introduce ChainUQ, a reasoning consistency-aware uncertainty quantification framework for LLMs. ChainUQ consists of two key technical components: an alignment-aware lightweight UQ module that estimates a raw intrinsic model confidence score from frozen features aligned to the final conclusion, and a reasoning consistency-aware calibrator that refines this score using reasoning-chain consistency evidence. Evaluations across diverse in-distribution and out-of-distribution benchmarks show that ChainUQ consistently improves response-level uncertainty estimation, achieving an average 3.1% relative gain in AUROC and up to 45.0% relative reduction in ECE, and can be directly transferred to new settings without additional fine-tuning.
