2 code implementations • IJCNLP 2019 • Robin Jia, aditi raghunathan, Kerem Göksel, Percy Liang
We train the first models that are provably robust to all word substitutions in this family.
Data Augmentation Natural Language Inference +1