Maxwell Ashford. Facilitating Cross-Domain Reasoning Generalization through Conservative Offline Reinforcement Learning Leveraging Pre-Trained Large Language Model Representations. CIS 2026, 4.