Hindsight Task Relabelling: Experience Replay for Sparse Reward Meta-RL

Neural Information Processing Systems 

Reinforcement learning (RL) has seen tremendous success applied to challenging games (Mnih et al.,