Interesting Problems

In your opinion, what are some of the most interesting/relevant open questions in RL right now? In any topic like inverse RL, imitation learning, model-based RL or more frontier lab focused like model-free, deep-RL, or RLHF-related questions.