Reward extrapolationIterative extrapolationKnowledge transferTrajectory-ranked reward extrapolation (T-REX) provides a general framework to infer users' intentions from sub-optimal demonstrations. However, it becomes inflexible when encountering multi-agent......