NeurIPS 2020

Critic Regularized Regression

Meta Review

This paper proposes a simple yet effective method by filtering off-distribution actions in the domain of offline RL. During the review process, some concerns were raised regarding the novelty of the paper (the method seems to be simple yet very effective, and there are not many deep theory or significantly new ideas). Correspondingly, there have been a long list of discussions on the paper. As a result, the reviewers believe that this kind of work also has its value, especially for practical applications (offline RL). Some constructive feedback were also provided for the authors to consider when writing the final version of the paper. BTW, the rebuttal is quite good and changed the mind of some reviewers.