Our work on sample-efficient adaptation of reward models for preference-based RL has been accepted at ICRA (link)!