Abstract: Offline reinforcement learning (RL) learns policies from fixed-size datasets without interacting with the environment, while multi-agent reinforcement learning (MARL) faces challenges from ...