Questions for the seminar paper "Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models".
-----------------------------------------------------------------------------------------------
Please send your answers to: mittal@cs.uni-freiburg.de

1) What is the motivation of learning in a reward-free offline regime? (2-3 sentences)
2) Under which conditions is model-based RL most effective? (2 sentences)
3) What is the role of JEPA in the proposed PLDM method? (1-2 sentences)