Provably Efficient Offline Reinforcement Learning in Regular Decision Processes | Read Paper on Bytez