We study offline multitask representation learning in reinforcement learning (RL), where a
learner is provided with an offline dataset from different tasks that share a common …