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.
@article{arxiv.2602.14518,
title = {Diagnosing Knowledge Conflict in Multimodal Long-Chain Reasoning},
author = {Jing Tang and Kun Wang and Haolang Lu and Hongjin Chen and KaiTao Chen and Zhongxiang Sun and Qiankun Li and Lingjuan Lyu and Guoshun Nan and Zhigang Zeng},
journal= {arXiv preprint arXiv:2602.14518},
year = {2026}
}