NeurIPS 2020

Sparse Graphical Memory for Robust Planning

Meta Review

Two referees support accept, one indicates reject. Rebuttal disputes R3's point about the usefulness of the central theorem and clarifies the comparison to the baselines. I find that the additional information provided in the rebuttal overcomes most of R3's concerns and recommend acceptance of the paper. However, I would like the authors to add their new theoretical insights to the paper, clarify performance differences to those presented in the original papers on the baselines and clarify states vs observations to make the paper more readable.