This repository will contain the code for our EACL 2021 paper:
Adv-OLM: Generating Textual Adversaries via OLM [Link to Paper]
- For running in your system:
conda create -n advolm python=3.6
conda activate advolm
pip install -r requirements.txt- For running on colab:
pip install -r requirements.txtpython attack_main.py --recipe advolm --batch-size 8 --num-examples 5 --model bert-base-uncased-imdb- For running other attacks like Textfooler, Bae, Pwws, etc., use TextAttack
@misc{malik2021advolm,
title={Adv-OLM: Generating Textual Adversaries via OLM},
author={Vijit Malik and Ashwani Bhat and Ashutosh Modi},
year={2021},
eprint={2101.08523},
archivePrefix={arXiv},
primaryClass={cs.CL}
}