Supervised Reward Inference

Will Schwarzer, Jordan Schneider, Philip S. Thomas +1

Existing approaches to reward inference typically assume that humans provide demonstrations according to specific behavior models. However, humans often indicate their goals throug…