STIR$^2$: Reward Relabelling for combined Reinforcement and Imitation Learning on sparse-reward tasks

Open in new window