重新审视基于有效秩的不确定性幻觉检测
Revisiting Hallucination Detection with Effective Rank-based Uncertainty
AI总结：
本文提出一种基于有效秩的不确定性量化方法，通过分析多输出多层的隐藏状态谱特征来检测大语言模型幻觉，无需额外模块，兼具理论依据与实验验证，实现稳健的幻觉检测。
摘要：
在大语言模型（LLM）中检测幻觉仍然是其可信部署面临的基本挑战。超越基于不确定性的传统幻觉检测框架，我们提出了一种简单而强大的方法，通过测量来自多个模型输出和不同层的隐藏状态的有效秩来量化不确定性。基于表示的谱分析，我们的方法通过语义变化为模型的内部推理过程提供可解释的见解，同时无需额外知识或附加模块，从而兼具理论优雅性和实际效率。此外，我们从理论上证明了在内部（单个响应的表示）和外部（不同响应）两个层面量化不确定性的必要性，为利用LLM不同层和不同响应的表示来检测幻觉提供了理论依据。大量实验表明，我们的方法能有效检测幻觉，并在各种场景下表现出稳健的泛化能力，为LLM真实性的幻觉检测贡献了新范式。
英文摘要：
Detecting hallucinations in large language models (LLMs) remains a fundamental challenge for their trustworthy deployment. Going beyond basic uncertainty-driven hallucination detection frameworks, we propose a simple yet powerful method that quantifies uncertainty by measuring the effective rank of hidden states derived from multiple model outputs and different layers. Grounded in the spectral analysis of representations, our approach provides interpretable insights into the model's internal reasoning process through semantic variations, while requiring no extra knowledge or additional modules, thus offering a combination of theoretical elegance and practical efficiency. Meanwhile, we theoretically demonstrate the necessity of quantifying uncertainty both internally (representations of a single response) and externally (different responses), providing a justification for using representations among different layers and responses from LLMs to detect hallucinations. Extensive experiments demonstrate that our method effectively detects hallucinations and generalizes robustly across various scenarios, contributing to a new paradigm of hallucination detection for LLM truthfulness.
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。
