Targeted Active Preference Learning for Underspecified Rewards
This work proposes active preference learning techniques to address poorly supervised features in underspecified reward models.
This work proposes active preference learning techniques to address poorly supervised features in underspecified reward models.