One Frozen Simulator Is Not Enough: Simulator Collapse in Multi-Agent RL

Part of

Advanced Inverse Reward And Prompt Engineering