This is a reinforcement learning framework designed to enhance emotional and cognitive reasoning in compact language models (LLMs). Grounded in psychological theory and dual-system reasoning, our method addresses key challenges in this emerging field: limited high-quality data, task-specific reasoning mismatch, and poor generalization in small models.

This repository accompanies our paper:

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

To run the code on sample tasks:

bash Stimuli2Minds/examples/start/bilateral.sh

This script launches the training pipeline using our Bilateral Reinforcement Learning (BR) framework with trajectory cache and reward shaping.