Contrastive Learning Tames Out-of-Distribution Actions in Offline Reinforcement Learning
Researchers at Hanyang University have developed TACCO, a contrastive learning method that explicitly identifies and suppresses out-of-distribution actions to stabilize ...
