Reflection Anchors for Propagation-Aware Visual Retention in Long-Chain Multimodal Reasoning

ArXi:2605.09614v1 Announce Type: new Long chain-of-thought (CoT) reasoning improves large vision--language models, but visual information often fades during generation, limiting long-horizon multimodal reasoning. Existing methods either re-inject vision at inference or train policies for stronger grounding, but where to intervene relies on perception heuristics rather than principled gain analysis, and how local visual influence propagates remains implicit.