MIRAGE：评估 MLLM 多模态推理链中的幻觉
MIRAGE: Assessing Hallucination in Multimodal Reasoning Chains of MLLM
AI总结：
针对多模态大语言模型中难以区分感知与推理幻觉的问题，提出 MIRAGE 基准隔离推理幻觉，并设计 RAG-RL 方法结合课程强化微调与协作提示推理，有效减少了基础模型的逻辑幻觉。
摘要：
多模态大语言模型（MLLMs）中的多模态幻觉限制了 MLLMs 的正确性。然而，多模态幻觉是多源的，并源于多种原因。现有的基准未能充分区分感知引起的幻觉和推理引起的幻觉。这一失败构成了一个重大问题，并阻碍了 MLLMs 内多模态推理失败的诊断。为了解决这个问题，我们提出了 MIRAGE 基准，它通过构建问题来隔离推理幻觉，在这些问题的输入图像中，MLLMs 能够正确感知，但推理错误依然存在。MIRAGE 引入了多粒度评估指标：准确率、事实性和用于幻觉量化的 LLMs 幻觉分数。我们的分析显示：（1）模型规模、数据规模和训练阶段显著影响逻辑、捏造和事实性幻觉的程度；（2）当前的 MLLMs 对因误解空间关系引起的空间幻觉没有显示出有效的改善，表明其视觉推理能力有限；（3）问题类型与不同的幻觉模式相关，突出了针对性的挑战和潜在的缓解策略。为了应对这些挑战，我们提出了 RAG-RL 方法，该方法结合了课程强化微调，通过逐步降低学习难度来鼓励模型生成逻辑一致的推理链，并结合协作提示推理来降低推理复杂性。RAG-RL 在 MIRAGE 上建立了一个基线，并减少了原始基础模型中的逻辑幻觉。
英文摘要：
Multimodal hallucination in multimodal large language models (MLLMs) restricts the correctness of MLLMs. However, multimodal hallucinations are multi-sourced and arise from diverse causes. Existing benchmarks fail to adequately distinguish between perception-induced hallucinations and reasoning-induced hallucinations. This failure constitutes a significant issue and hinders the diagnosis of multimodal reasoning failures within MLLMs. To address this, we propose the {\dataset} benchmark, which isolates reasoning hallucinations by constructing questions where input images are correctly perceived by MLLMs yet reasoning errors persist. {\dataset} introduces multi-granular evaluation metrics: accuracy, factuality, and LLMs hallucination score for hallucination quantification. Our analysis reveals that (1) the model scale, data scale, and training stages significantly affect the degree of logical, fabrication, and factual hallucinations; (2) current MLLMs show no effective improvement on spatial hallucinations caused by misinterpreted spatial relationships, indicating their limited visual reasoning capabilities; and (3) question types correlate with distinct hallucination patterns, highlighting targeted challenges and potential mitigation strategies. To address these challenges, we propose {\method}, a method that combines curriculum reinforcement fine-tuning to encourage models to generate logic-consistent reasoning chains by stepwise reducing learning difficulty, and collaborative hint inference to reduce reasoning complexity. {\method} establishes a baseline on {\dataset}, and reduces the logical hallucinations in original base models.
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。
