Weighted model estimation for offline model-based reinforcement learning

This paper discusses model estimation in offline model-based reinforcement learning (MBRL), which is important for subsequent policy improvement using an estimated model.