NeurIPS 2020

Regularizing Black-box Models for Improved Interpretability

Meta Review

The reviewers agree that the problem paper is trying to solve is quite important, and through very strong experiments, show that the approach is quite effective. The biggest concern that was brought up in the discussion was the lack of novelty. The reviewers strongly urge the authors to focus more on the strength of their contributions, in particular the evaluation, as opposed to trying to differentiate from work they are clearly building upon.