大推理模型中幻觉的检测与缓解：一种机制视角
Detection and Mitigation of Hallucination in Large Reasoning Models: A Mechanistic Perspective
AI总结：
本研究从机制视角探究大推理模型的推理幻觉问题，提出推理得分量化推理深度，构建RHD检测框架实现多领域最优检测效果，还提出GRPO-R强化学习算法缓解幻觉，兼具理论泛化保证与实验效果提升。
摘要：
大推理模型（LRMs）在多步推理任务中展现出了令人瞩目的能力。然而，伴随这些进展，一种更具欺骗性的模型错误应运而生——推理幻觉，即逻辑连贯但事实有误的推理链会得出看似可信实则错误的结论。与传统幻觉不同，这类错误嵌入在结构化推理过程中，更难被检测到，潜在危害也更大。\n在本研究中，我们从机制视角探究推理幻觉问题。我们提出了“推理得分”（Reasoning Score），通过测量LRM后期层投影到词汇表空间得到的logits之间的差异，来量化推理深度，有效区分浅层模式匹配与真正的深度推理。利用该得分，我们在ReTruthQA数据集上开展了深入分析，识别出两种关键的推理幻觉模式：推理深度的早期波动，以及回溯到存在缺陷的先前步骤的错误回溯。\n基于这些发现，我们构建了推理幻觉检测（RHD）框架，该框架在多个领域均取得了当前最优性能。为缓解推理幻觉，我们进一步提出了GRPO-R——一种增强型强化学习算法，它通过基于势函数的塑形引入步骤级的深度推理奖励。我们的理论分析证明了该方法具备更强的泛化保证，实验结果也表明其提升了推理质量，降低了幻觉发生率。
英文摘要：
Large Reasoning Models (LRMs) have shown impressive capabilities in multi-step reasoning tasks. However, alongside these successes, a more deceptive form of model error has emerged--Reasoning Hallucination--where logically coherent but factually incorrect reasoning traces lead to persuasive yet faulty conclusions. Unlike traditional hallucinations, these errors are embedded within structured reasoning, making them more difficult to detect and potentially more harmful. In this work, we investigate reasoning hallucinations from a mechanistic perspective. We propose the Reasoning Score, which quantifies the depth of reasoning by measuring the divergence between logits obtained from projecting late layers of LRMs to the vocabulary space, effectively distinguishing shallow pattern-matching from genuine deep reasoning. Using this score, we conduct an in-depth analysis on the ReTruthQA dataset and identify two key reasoning hallucination patterns: early-stage fluctuation in reasoning depth and incorrect backtracking to flawed prior steps. These insights motivate our Reasoning Hallucination Detection (RHD) framework, which achieves state-of-the-art performance across multiple domains. To mitigate reasoning hallucinations, we further introduce GRPO-R, an enhanced reinforcement learning algorithm that incorporates step-level deep reasoning rewards via potential-based shaping. Our theoretical analysis establishes stronger generalization guarantees, and experiments demonstrate improved reasoning quality and reduced hallucination rates.
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。
