← Return to Article Details Reward Redistribution as Align-RUDDER: Learning from a Few Demonstrations