RESEARCH

Omni-Perception Policy Optimization for Multimodal Emotion Reasoning

ArXiv cs.AI · Thu, 25 Jun 2026 04:00:00 GMT

arXiv:2606.25325v1 Announce Type: new Abstract: We find that current emotion-oriented Omni-MLLMs still lack reliable omni-modal perception: they (i) underutilize multimodal cues in their reasoning trajectories and (ii) exhibit unfaithful behavior, often hallucinating modality-spe

Read original source Discuss with SiMON