@misc{indiciae9406f9858818,
  title = {Defending Pre-trained Language Models from Adversarial Word Substitutions Without Performance Sacrifice},
  author = {Rongzhou Bao and Jiayi Wang and Hai Zhao},
  year = {2021},
  url = {https://arxiv.org/abs/2105.14553},
  note = {Source identifier: 2105.14553}
}