Code for Toward Stronger Textual Attack Detectors

TODO : update before public release of code +scripts

Requirements

We provide A SAMPLE OF Stakout in DATASET/ due to size constraints of openreview.

The models we used are available in the HuggingFace Hub: https://huggingface.co/textattack. To generate the attacks we used TextAttack https://github.com/QData/TextAttack.

An example of bash example is given in:

>> sh eval_adv.sh

>> python run_baseline_gpt2.py

An example of bash example is given in:

>> sh get_distance.sh

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
DATASET		DATASET
README.md		README.md
data_depth.py		data_depth.py
eval_adv.sh		eval_adv.sh
eval_pretrained_model_attack.py		eval_pretrained_model_attack.py
evaluation.py		evaluation.py
get_distance.sh		get_distance.sh
get_distances.py		get_distances.py
metrics.py		metrics.py
model.py		model.py
run_baseline_gpt2.py		run_baseline_gpt2.py
utils.py		utils.py
utils_evaluation.py		utils_evaluation.py