Omni-Perception Policy Optimization for Multimodal Emotion Reasoning

[Submitted on 24 Jun 2026 (v1), last revised 20 Jul 2026 (this version, v2)]

View PDF HTML (experimental)

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-specific statements from other modalities. Building on these insights, we propose OPPO (Omni-Perception Policy Optimization), a reinforcement learning framework that explicitly optimizes multimodal perception. First, an Omni-Perception Reward decomposes ground-truth reasoning into fine-grained visual, acoustic, and emotion cues and rewards trajectories that semantically recover these cues. Second, an Omni-Perception Loss compares the policy under full and unimodally masked inputs, applying a KL penalty only to modality-specific evidence tokens to suppress cross-modal hallucination. We further introduce MEP-Bench, a diagnostic benchmark that quantifies utilization and faithfulness. Experiments show that OPPO achieves state-of-the-art performance on MER-UniBench and MME-Emotion, while substantially improving utilization and faithfulness scores on MEP-Bench, highlighting the importance of sufficient and faithful omni perception for multimodal emotion reasoning.

Submission history

From: Zhiyuan Han [view email]
[v1] Wed, 24 Jun 2026 02:43:26 UTC (7,878 KB)
[v2] Mon, 20 Jul 2026 13:20:39 UTC (7,879 KB)