Mitigating the Inconsistency Between Word Saliency and Model Confidence with Pathological Contrastive Training

Pengwei Zhan, Yang Wu, Shaolei Zhou, Yunjian Zhang, Liming Wang