MER-R1: Multimodal Emotion Reasoning via Slow-Fast Thinking Synergy

2026-06-25

AI Summary

This work addresses the challenge in multimodal emotion recognition that explicit reasoning, while enhancing interpretability, often struggles to simultaneously achieve high precision and recall. To overcome this trade-off, the paper proposes MER-R1, a novel framework that unifies fast-thinking (high-recall) and slow-thinking (high-precision) strategies at the optimization level. Through a dual-objective decoupling mechanism and confidence calibration, MER-R1 enables synergistic rather than compromised integration of both reasoning modes, with theoretical analysis demonstrating its ability to mitigate variance-induced interference. Built upon a reinforcement learning–based multimodal large language model architecture, MER-R1 achieves state-of-the-art performance on the MER-UniBench and MME-Emotion benchmarks, validating that the proposed collaborative reasoning mechanism effectively enhances emotion recognition accuracy.

Abstract

We find that explicit reasoning does not necessarily translate into better multimodal emotion recognition (MER) accuracy, even though it makes predictions more interpretable. Specifically, for reasoning-based MLLMs, fast thinking by triggering direct answers often outperforms slow thinking after deliberative reasoning. Our empirical analyses show that fast thinking improves recall with broader and more confident predictions, whereas slow thinking favors precision through conservative filtering of incorrect categories. Building on these insights, we propose MER-R1, a reinforcement learning framework that turns slow-fast complementarity into explicit optimization. Dual-objective disentanglement separates recall and precision into two optimization signals, allowing them to be jointly optimized rather than traded off against each other. Slow-fast confidence calibration further aligns the final slow-thinking answer with fast-thinking intuition, strengthening correct emotions while suppressing incorrect ones. In this way, MER-R1 unifies the recall-oriented intuition of fast thinking with the precision-oriented selectivity of slow thinking. We further provide theoretical justification for this synergy, showing that it mitigates variance-induced interference during optimization. Extensive experiments on MER-UniBench and MME-Emotion show that MER-R1 achieves state-of-the-art performance and makes reasoning genuinely benefit emotion recognition.

Problem

• Multimodal Emotion Recognition
• Slow-Fast Thinking
• Recall-Precision Trade-off
• Emotion Reasoning
• MLLMs

Innovation

• slow-fast thinking
• multimodal emotion recognition
• reinforcement learning
• dual-objective disentanglement
• confidence calibration