Rewriting History with Inverse RL:
      Hindsight Inference for Policy Improvement

  Benjamin Eysenbachφθ Xinyang Gengψ         Sergey Levineψθ    Ruslan Salakhutdinovφ
       φ                    ψ           θ
        Carnegie Mellon University        UC Berkeley      Google Brain

                        Abstract
      Multi-task reinforcement learning (RL) aims to simultaneously learn policies for
      solving many tasks. Several prior works have found that relabeling past experience
    ...