通过分析推理过程的稳定性来检测大模型幻觉，效果显著提升。
Harnessing Reasoning Trajectories for Hallucination Detection via Answer-agreement Representation Shaping
- 通过扰动推理边界嵌入生成反事实答案，判断一致性
- 使一致答案对应的表示聚拢，不一致则分离，暴露幻觉风险
- 无需人工标注，可直接接入现有检测器使用
大型推理模型（LRMs）常生成看似连贯的长推理轨迹却给出错误答案，导致幻觉检测困难。尽管轨迹包含有用信号，但直接使用文本或原始隐藏状态进行检测易受形式变化影响，且容易过拟合表面模式而非答案真实性。本文提出答案一致性表示塑造（ARS），通过显式编码答案稳定性来学习利于检测的条件化表示。ARS通过微小的潜在干扰（具体为扰动轨迹边界嵌入）生成反事实答案，并以结果是否与原答案一致作为标签。随后学习将一致状态聚集、不一致状态分离的表示，从而揭示潜在的不稳定性，反映幻觉风险。所获嵌入可即插即用至现有基于嵌入的检测器中，训练过程无需人类标注。实验表明，ARS持续提升检测性能，显著优于强基线方法。代码已公开于：https://github.com/radiolab-ntu/ars_icml2026。
原文摘要 · Abstract (English)
Large reasoning models (LRMs) often generate long, seemingly coherent reasoning traces yet still produce incorrect answers, making hallucination detection challenging. Although trajectories contain useful signals, directly using trace text or vanilla hidden states for detection is brittle: traces vary in form and detectors can overfit to superficial patterns rather than answer validity. We introduce Answer-agreement Representation Shaping (ARS), which learns detection-friendly trace-conditioned representations by explicitly encoding answer stability. ARS generates counterfactual answers through small latent interventions, specifically, perturbing the trace-boundary embedding, and labels each perturbation by whether the resulting answer agrees with the original. It then learns representations that bring answer-agreeing states together and separate answer-disagreeing ones, exposing latent instability indicative of hallucination risk. The shaped embeddings are plug-and-play with existing embedding-based detectors and require no human annotations during training. Experiments demonstrate that ARS consistently improves detection and achieves substantial gains over strong baselines. Code is available at: https://github.com/radiolab-ntu/ars_icml2026.
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。