Abductive natural language inference by interactive model with structural loss

Linhao Li, Ao Wang, Ming Xu, Yongfeng Dong, Xin Li

Abstract: Highlights•For α<math><mi is="true">α</mi></math>NLI task, we regroup instead of ranking all hypotheses.•We design a softmax focal loss for each group and combine them into a joint loss.•we design an information interaction layer that increases the AUC by about 5%.