arXiv 2505.200452026-06-18cs.CL

基于不确定性感知注意力头的高效大语言模型幻觉检测

Efficient Hallucination Detection for LLMs Using Uncertainty-Aware Attention Heads

Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)（莫扎德人工智能大学）；ETH Zurich（苏黎世联邦理工学院）；Independent Researcher（独立研究者）；Applied AI Institute（应用人工智能研究所）

Artem Vazhentsev, Lyudmila Rvanova, Gleb Kuzmin, Ekaterina Fadeeva, Ivan Lazichny, Alexander Panchenko, Maxim Panov, Mrinmaya Sachan, Preslav Nakov, Timothy Bal… 展开作者

Artem Vazhentsev, Lyudmila Rvanova, Gleb Kuzmin, Ekaterina Fadeeva, Ivan Lazichny, Alexander Panchenko, Maxim Panov, Mrinmaya Sachan, Preslav Nakov, Timothy Baldwin, Artem Shelmanov

发表信息：: Proceedings of the 43rd International Conference on Machine Learning (ICML), Seoul, South Korea, 2026

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

提出RAUQ框架，利用不确定性感知注意力头与令牌级置信度，通过单次前向传递实现无监督、高效的序列级幻觉检测，在12个数据集上优于现有方法且额外计算少于1%。

摘要：

尽管大型语言模型（LLM）已经变得非常强大，但它们仍然容易出现事实性错误，通常称为“幻觉”。不确定性量化（UQ）为缓解这一问题提供了一种有前景的方法，但大多数现有方法计算量大且/或需要监督。在这项工作中，我们提出了基于循环注意力的不确定性量化（RAUQ），这是一种无监督且高效的幻觉识别框架。该方法利用了Transformer注意力行为的一个观察：当生成错误信息时，某些“不确定性感知”注意力头倾向于减少对前驱令牌的关注。RAUQ自动检测这些注意力头，并以循环方式将其激活模式与令牌级置信度度量相结合，仅通过一次前向传递即可生成序列级不确定性估计。通过在涵盖问答、摘要和翻译的十二个数据集上对九个不同LLM进行的实验，我们表明RAUQ始终优于最先进的UQ基线。重要的是，它产生的开销极小，所需的额外计算不到1%。由于它既不需要标记数据也不需要广泛的参数调整，RAUQ可作为白盒LLM中实时幻觉检测的轻量级即插即用解决方案。

英文摘要：

While large language models (LLMs) have become highly capable, they remain prone to factual inaccuracies, commonly referred to as "hallucinations." Uncertainty quantification (UQ) offers a promising way to mitigate this issue, but most existing methods are computationally intensive and/or require supervision. In this work, we propose Recurrent Attention-based Uncertainty Quantification (RAUQ), an unsupervised and efficient framework for identifying hallucinations. The method leverages an observation about transformer attention behavior: when incorrect information is generated, certain "uncertainty-aware" attention heads tend to reduce their focus on preceding tokens. RAUQ automatically detects these attention heads and combines their activation patterns with token-level confidence measures in a recurrent scheme, producing a sequence-level uncertainty estimate in just a single forward pass. Through experiments on twelve datasets spanning question answering, summarization, and translation across nine different LLMs, we show that RAUQ consistently outperforms state-of-the-art UQ baselines. Importantly, it incurs minimal overhead, requiring less than 1\% additional computation. Since it requires neither labeled data nor extensive parameter tuning, RAUQ serves as a lightweight, plug-and-play solution for real-time hallucination detection in white-box LLMs.

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。