Certified Robustness

Robustness-Aware Word Embedding Improves Certified Robustness to Adversarial Word Substitutions
Natural Language Processing (NLP) models have gained great success on clean texts, but they are known to be vulnerable to adversarial …
Yibin Wang, Yichen Yang, Di He, Kun He
PDF Cite Code ACL Anthology
Robustness-Aware Word Embedding Improves Certified Robustness to Adversarial Word Substitutions