Phantivia / T-PGD

[Findings of ACL 2023] Bridge the Gap Between CV and NLP! A Optimization-based Textual Adversarial Attack Framework.

T-PGD

Code and data of the Findings of ACL 2023 paper "Bridge the Gap Between CV and NLP! A Gradient-based Textual Adversarial Attack Framework"

How to run

Please check T-PGD/LaunchTPGD.ipynb to see about the details of hyperparameters and we have all the commands to run our main experiments there.

Set up Metric

Before running the experiments, please download the USE-4 model from https://tfhub.dev/google/universal-sentence-encoder/4 and set the path variable in utils/Metric.py

Requirements

The main packages we used in this project are listed below:

python==3.10.0
torch==1.13.1
transformers==4.29.0
tensorflow==2.12.0
tensorflow_hub==0.13.0
language-tool-python==2.7.1

Citation

Please kindly cite our paper:

@inproceedings{yuan-etal-2023-bridge,
    title = "Bridge the Gap Between {CV} and {NLP}! A Gradient-based Textual Adversarial Attack Framework",
    author = "Yuan, Lifan  and
      Zhang, YiChi  and
      Chen, Yangyi  and
      Wei, Wei",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
    year = "2023",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.findings-acl.446"
}