Adversarial Text Generation by Search and Learning

Our 7 datasets are:

Required packages are listed in the requirements.txt file:

pip install -r requirements.txt

• Run the following code to generate the adversaries:

python command_dataset_model.py

Here we explain each required argument in details:

• --data_set: The path to the dataset. We used the public datasets datasets.
• --victim: Name of the target model such as ''bert''. trained model parameters.
• --bm_model: Our generated finely tuned BM_Model,we shared the trained model parameters.
• --train_classification: Used to generate victim models.
• --model: All core model codes of ATGSL.

In case someone may want to use our generated adversary results towards the benchmark data directly, here it is.