NeurIPS 2020

Hard Negative Mixing for Contrastive Learning

Meta Review

All the reviewers agree this paper is strong. Due to the popularity of contrastive learning and the ideas in this paper, the paper will be timely at NeurIPS. The reviewers also praised the strong empirical results, especially over established baselines. The careful analysis and use of hard negative mining leads to these strong results, as noted by the reviewers.