Emotion-o1: Adaptive Long Reasoning for Emotion Understanding in LLMs

Changhao Song, Yazhou Zhang, Hui Gao, Kaiyun Huang, Peng Zhang

Keywords: Large Language Model, Affective computing, Long Chain-of-Thought, Adaptive Length

Abstract: Long chain-of-thought (CoT) reasoning has shown great promise in enhancing the emotion understanding performance of large language models (LLMs). However, current fixed-length CoT methods struggle to balance reasoning depth and efficiency. Simple tasks (e.g., sentiment classification) are over-reasoned, while complex tasks (e.g., sarcasm understanding) lack depth. To fill this gap, we present Emotion-o1, an adaptive CoT framework that dynamically adjusts reasoning length based on task complexity. Emotion-o1 is trained by distilling adaptive CoT patterns from a large reasoning model (LRM), followed by supervised fine-tuning and reinforcement learning with a four-part reward targeting accuracy, brevity, structure, and redundancy. Experimental results on four emotion tasks highlight: (1) Emotion-o1 demonstrates significant improvements over its backbone, with F1 score increases of 11\%↑(Sentiment), 14\%↑(Emotion), 18\%↑(Humor), and 27\%↑(Sarcasm). (2) In sentiment and emotion tasks, our 8B model demonstrates superior performance against SoTA LLMs, outperforming Grok‑3 by 2.1\% in sentiment and within 1\% of OpenAI‑o1 in emotion. (3) The framework maintains accuracy while reducing reasoning length by 83\% compared to OpenAI-o1, demonstrating effective precision-efficiency optimization. From a lower-cost perspective, the framework also empowers SLMs to achieve reasoning capabilities comparable to larger ones.

Primary Area: applications to computer vision, audio, language, and other modalities

Submission Number: 8326