通过解析模型内部信号，精准识别RAG生成中的幻觉问题
InterpDetect: Interpretable Signals for Detecting Hallucinations in Retrieval-Augmented Generation
- 分析RAG幻觉成因：后期前馈层过度注入参数化知识
- 提出外部上下文与参数知识得分，训练回归分类器检测幻觉
- 小模型信号可泛化至大模型，适合评估RAG系统可靠性
检索增强生成（RAG）通过引入外部知识缓解幻觉，但模型常生成与检索内容不一致的输出。准确检测幻觉需分离外部上下文与参数化知识的贡献，而现有方法常将其混淆。我们研究RAG幻觉机制，发现其源于后期前馈网络（FFN）模块过度将参数化知识注入残差流。为此，我们提出基于外部上下文得分与参数知识得分的可解释检测方法。以Qwen3-0.6b为实验模型，计算各层与注意力头的得分，并训练回归分类器预测幻觉。方法在GPT-5、GPT-4.1及RAGAS、TruLens、RefChecker等基线中评估。结果显示，基于Qwen3-0.6b信号训练的分类器可有效泛化至GPT-4.1-mini生成结果，证明代理模型评估的潜力。研究表明，机制性信号是高效且可泛化的幻觉检测指标。
原文摘要 · Abstract (English)
Retrieval-Augmented Generation (RAG) integrates external knowledge to mitigate hallucinations, yet models often generate outputs inconsistent with retrieved content. Accurate hallucination detection requires disentangling the contributions of external context and parametric knowledge, which prior methods typically conflate. We investigate the mechanisms underlying RAG hallucinations and find they arise when later-layer FFN modules disproportionately inject parametric knowledge into the residual stream. To address this, we explore a mechanistic detection approach based on external context scores and parametric knowledge scores. Using Qwen3-0.6b, we compute these scores across layers and attention heads and train regression-based classifiers to predict hallucinations. Our method is evaluated against state-of-the-art LLMs (GPT-5, GPT-4.1) and detection baselines (RAGAS, TruLens, RefChecker). Furthermore, classifiers trained on Qwen3-0.6b signals generalize to GPT-4.1-mini responses, demonstrating the potential of proxy-model evaluation. Our results highlight mechanistic signals as efficient, generalizable predictors for hallucination detection in RAG systems.
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。