2 datasets found

ObjectNav revisited: On evaluation of embodied agents navigating to objects.

POLIFORMER is a transformer-based policy model trained with on-policy RL in the AI2-THOR simulator at scale, that can effectively navigate in the real world without any adaptation.