NeurIPS 2020

Learning Robust Decision Policies from Observational Data

Meta Review

The reviewers and myself agree that the paper provides a strong conceptual contribution in analyzing robust off policy learning. There is some criticism on the presentation and the experimental part: 1) the authors are strongly encouraged to include the new stronger baseline that is described in their rebuttal, 2) the authors are strongly encouraged to discuss the benefits of their algorithm in settings with poor overlap. Despite these drawbacks the conceptual contribution of the work seems strong enough to merit acceptance.