Attention at Rest Stays at Rest: Breaking Visual Inertia for Cognitive Hallucination Mitigation

ArXi:2604.01989v1 Announce Type: new Like a body at rest that stays at rest, we find that visual attention in multimodal large language models (MLLMs) exhibits pronounced inertia, remaining largely static once settled during early decoding steps and failing to the compositional understanding required for cognitive inference. While existing hallucination mitigation methods mainly target perceptual hallucinations concerning object existence or attributes, they remain inadequate for such cognitive hallucinations that require inter-object relational deduction.