NeurIPS 2020

RD$^2$: Reward Decomposition with Representation Decomposition


Meta Review

The authors propose an automatic reward decomposition method that allows better credit assignment. The reviewers agree that the approach is interesting and intuitive, and experimental results are positive and include interesting games. The rebuttal was very helpful in clarifying questions raised. Please make sure that you include these clarifications as well as the extra results in the final version of the paper.