Omni-Perception Policy Optimization for
Multimodal Emotion Reasoning

Zhiyuan Han1,2,4, Beier Zhu1, Wenwen Tong2, Pengyang Shao3, Peipei Song1, Xinyi Wang1
Jiangnan Chen2, Lewei Lu2, Xun Yang1
1University of Science and Technology of China, 2SenseTime Research
3National University of Singapore, 4Institute of Artificial Intelligence, Hefei Comprehensive National Science Center
International Conference on Machine Learning (ICML) 2026
OPPO teaser showing utilization and faithfulness issues

OPPO improves multimodal emotion reasoning by encouraging models to use enough evidence and stay faithful to the modality that supports each claim.

Abstract

Multimodal emotion reasoning requires models to recognize emotions and explain them with evidence from visual behavior, acoustic patterns, and language context. Although recent Omni-MLLMs have shown promising reasoning ability, their explanations are not always perceptually reliable: a model may ignore useful fine-grained cues, or mention visual and acoustic evidence that is not actually supported by the input.

We identify two key perception problems behind this failure: insufficient utilization of multimodal evidence and unfaithful cross-modal hallucination. To address them, we propose Omni-Perception Policy Optimization (OPPO), a perception-aware optimization framework for emotion Omni-MLLMs.

OPPO introduces an Omni-Perception Reward to encourage generated reasoning to recover annotated video, audio, and emotion cues, and an Omni-Perception Loss to test whether modality-specific claims remain valid when video or audio evidence is masked. These two objectives jointly push the model to use enough evidence while keeping its explanations grounded in the correct modality.

We further build MEP-Bench, a diagnostic benchmark for measuring evidence utilization and perception faithfulness. Experiments on MEP-Bench and MER-UniBench show that improving perceptual grounding leads to more reliable reasoning and stronger multimodal emotion recognition performance.

Method Overview

OPPO optimizes multimodal emotion reasoning from two complementary perspectives: whether the model uses enough perceptual evidence, and whether each modality-specific claim is faithful to the modality that actually supports it. As shown in the overview figure, the upper branch builds an evidence-aware reward, while the lower branch introduces a modality-masking loss to suppress hallucinated visual or acoustic cues.

Omni-Perception Reward: Given human-annotated video, audio, and emotion cues, OPPO first constructs a multimodal evidence set. The generated reasoning is then split into clauses and aligned with these cues through an evidence-routing matrix. This reward encourages the model to recover sufficient emotion-relevant evidence during reasoning, instead of only optimizing for the final emotion label or response format.

Omni-Perception Loss: To make the reasoning faithful to its perceptual source, OPPO masks one modality at a time and compares the token-level distributions of modality-specific evidence before and after masking. If a visual or acoustic claim remains confident even when its supporting modality is removed, the loss penalizes it. This discourages cross-modal hallucination and keeps the explanation grounded in visible facial behavior, acoustic cues, and valid emotion evidence.

Method overview of the OPPO framework

Qualitative Comparison

Evidence Utilization

Qualitative utilization comparison between AffectGPT-R1 and OPPO

Visual Faithfulness

Qualitative video faithfulness comparison between AffectGPT-R1 and OPPO

Qualitative comparison: OPPO retrieves richer multimodal evidence during reasoning and avoids hallucinating modality-specific cues under masking. Compared with the baseline, the explanations are more tightly grounded in visible facial behavior, acoustic cues, and emotion-relevant evidence.

Quantitative Results

MEP-Bench Diagnostic Benchmark

Diagnostic results on MEP-Bench

MEP-Bench is derived from OV-MERD and evaluates whether reasoning uses human-annotated multimodal cues and whether visual or acoustic claims remain faithful under unimodal masking. It separates underutilization from cross-modal hallucination through evidence recall and masked perception accuracy.

Main Results on MER-UniBench

Main results on MER-UniBench

OPPO achieves the best mean score among compared Omni-MLLM baselines, with consistent improvements across sentiment analysis, basic emotion recognition, and fine-grained emotion recognition.

Main Results on MME-Emotion

Main results on MME-Emotion

OPPO achieves the best overall performance on MME-Emotion, improving task-level emotion reasoning and the mean recognition, reasoning, and CoT scores.

BibTeX

@inproceedings{han2026omniperception,
  title={Omni-Perception Policy Optimization for Multimodal Emotion Reasoning},
  author={Zhiyuan Han and Beier Zhu and Wenwen Tong and Pengyang Shao and Peipei Song and Xinyi Wang and Jiangnan Chen and Lewei Lu and Xun Yang},
  booktitle={Forty-third International Conference on Machine Learning},
  year={2026},
  url={https://openreview.net/forum?id=4dwe2wlxx5}
}