Fetching the paper…

Cal-QL: Calibrated Offline RL Pre-Training for Efficient Online Fine-Tuning · Around