Reviews: On Mixup Training: Improved Calibration and Predictive Uncertainty for Deep Neural Networks

This paper investigates the use of mixup to improve the calibration of neural nets. Neural nets are known to be poorly calibrated and this poses significant problems in several important applications. The reviewers found that this work provides compelling empirical evidence that mixup address this important problem. The concerns raised by the reviewers were sufficiently addressed by the rebuttal. This work would be of interest to the NeurIPS community.