TY  - RPRT
TI  - Defending Pre-trained Language Models from Adversarial Word Substitutions Without Performance Sacrifice
AU  - Rongzhou Bao
AU  - Jiayi Wang
AU  - Hai Zhao
PY  - 2021
UR  - https://arxiv.org/abs/2105.14553
ID  - 2105.14553
ER  -