From Stimuli to Minds: Enhancing Psychological Reasoning in LLMs via Bilateral Reinforcement Learning

Large Language Models show promise in emotion understanding, social reasoning, and empathy, yet they struggle with psychologically grounded tasks that require inferring implicit mental states in context-rich, ambiguous settings. These limitations arise from the absence of theory-aligned supervision and the difficulty of capturing nuanced mental processes in real-world narratives. To address this gap, we leverage expert-labeled, psychologically rich scenarios and propose a trajectory-aware reinforcement learning framework that explicitly imitates expert psychological thought patterns. By integrating real-world stimuli with structured reasoning guidance, our approach enables compact models to internalize social-cognitive principles, perform nuanced psychological inference, and support continual self-improvement. Experiments across multiple benchmarks demonstrate that our models achieve expert-level interpretive capabilities, exhibiting strong out-of-distribution generalization across diverse psychologically tasks. Our code is publicly available at https://github.com/Githubuseryf/Stimuli2Minds . * These authors contributed equally. Mrs. Li's son, Xiao Qiang, is in college now, he tells her that he is the captain of the school football team. This achievement is high because the football team is very popular in school. Question: How does Mrs. Li react when Xiao Qiang is chosen as the football team captain? Mrs. Li's son, Xiaoqiang, goes to college and tells her that he is the captain of the school football team. This achievement is high because the football team is very popular in school. But Mrs. Li thinks her son may face pressure in balancing study and football activities. Xiaoqiang tells Mrs. Li that he is the captain of the football team. Question: What emotion does Mrs. Li possibly show? Mrs. Li would likely be very proud of her son for being chosen as the football team captain