用激活值谱特征实时检测大模型幻觉和分布外问题
EigenTrack: Spectral Activation Feature Tracking for Hallucination and Out-of-Distribution Detection in LLMs and VLMs
- 通过隐藏层激活的谱几何特性构建全局动态签名
- 仅需一次前向传播即可实时检测幻觉与分布外偏差
- 可解释且兼顾准确率与延迟，适合部署于生产环境
大语言模型虽具广泛用途，但仍易产生幻觉和分布外（OOD）错误。我们提出EigenTrack，一种可解释的实时检测器，利用隐藏激活的谱几何特性——即模型动态的紧凑全局签名。通过流式传输协方差谱统计量（如熵、特征值间距、与随机基线的KL散度）至轻量级递归分类器，EigenTrack 能追踪表示结构的时序变化，在表面错误出现前识别幻觉与分布外漂移。相比黑盒与灰盒方法，它只需单次前向传播，无需重采样；相比现有白盒检测器，它保留时序上下文，聚合全局信号，并提供可解释的精度-延迟权衡。
原文摘要 · Abstract (English)
Large language models (LLMs) offer broad utility but remain prone to hallucination and out-of-distribution (OOD) errors. We propose EigenTrack, an interpretable real-time detector that uses the spectral geometry of hidden activations, a compact global signature of model dynamics. By streaming covariance-spectrum statistics such as entropy, eigenvalue gaps, and KL divergence from random baselines into a lightweight recurrent classifier, EigenTrack tracks temporal shifts in representation structure that signal hallucination and OOD drift before surface errors appear. Unlike black- and grey-box methods, it needs only a single forward pass without resampling. Unlike existing white-box detectors, it preserves temporal context, aggregates global signals, and offers interpretable accuracy-latency trade-offs.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。