Representation Learning

VIP: Towards Universal Visual Reward and Representation via Value-Implicit Pre-Training