NeurIPS 2020

Self-Supervised Relationship Probing

Meta Review

Overall this is an interesting paper with very interesting ablations and experimental results. Authors also propose a new way of doing self-supervised learning. The results in this paper show that probing and the data augmentation both are useful. However, the writing of the paper requires some revision, at this stage the paper is a bit difficult to read which I think can be fixed with minor changes on the writing.