Image
MER-R1: Chinese Researchers Reveal AI Thinking Paradox — Slow Reasoning Can Harm Emotion Recognition
A new study from USTC and SenseTime researchers shows that when AI models "think" about emotions, they can perform worse than when answering instantly. The MER-R1 framework solves this paradox through clever synergy of both approaches, achieving record accuracy on the MER-UniBench and MME-Emotion benchmarks.