[2302.05372] Towards Minimax Optimality of Model-based Robust Reinforcement Learning