多模态长链推理中的知识冲突诊断
Diagnosing Knowledge Conflict in Multimodal Long-Chain Reasoning
AI总结：
本研究提出了一种统一的知识冲突概念，揭示了多模态长链推理中不同冲突类型的特征和处理机制，为诊断和控制推理失败提供了原理性方法。
摘要：
多模态大语言模型（MLLMs）在长链推理中常常在不同知识源提供冲突信号时失败。我们在此基础上提出了一种统一的知识冲突概念，区分输入层面的客观冲突与处理层面的有效冲突。通过探测内部表示，我们揭示了：（I）线性可分性：不同冲突类型被显式编码为线性可分的特征，而非纠缠；（II）深度定位：冲突信号集中在中到后期层，表明冲突编码有 distinct 的处理阶段；（III）层次一致性：沿轨迹聚合噪声的token级信号能稳健地恢复输入层面的冲突类型；（IV）方向不对称性：在冲突下强化模型的隐含源偏好比强制相反的源更容易。我们的发现为知识冲突下的多模态推理提供了机制层面的视角，并实现了长链推理失败的原理性诊断与控制。
英文摘要：
Multimodal large language models (MLLMs) in long chain-of-thought reasoning often fail when different knowledge sources provide conflicting signals. We formalize these failures under a unified notion of knowledge conflict, distinguishing input-level objective conflict from process-level effective conflict. Through probing internal representations, we reveal that: (I) Linear Separability: different conflict types are explicitly encoded as linearly separable features rather than entangled; (II) Depth Localization: conflict signals concentrate in mid-to-late layers, indicating a distinct processing stage for conflict encoding; (III) Hierarchical Consistency: aggregating noisy token-level signals along trajectories robustly recovers input-level conflict types; and (IV) Directional Asymmetry: reinforcing the model's implicit source preference under conflict is far easier than enforcing the opposite source. Our findings provide a mechanism-level view of multimodal reasoning under knowledge conflict and enable principled diagnosis and control of long-CoT failures.
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。
