EditBias: Debiasing Language Models via Model Editing

Setup

Environment

This codebase uses Python 3.9.18. Other versions may work as well.

Create an environment and install the dependencies:

$ conda create -n editbias python=3.9
$ conda activate editbias
(editbias) $ pip install -r requirements.txt

EditBias

• Training

   (editbias) $ bash scripts/gpt2-xl.sh >scripts/gpt2-xl.log 2>&1

  Record the path $p_1$ of the final parameters of the editor networks in the training log.

• Evaluation

  1. Set archive as $p_1$ in the evaluation script.

  2. Run bash scripts/gpt2-xl_eval.sh >scripts/gpt2-xl_eval.log 2>&1 and record the output path $p_2$ in the evaluation log.

  3. Set root as $p_2$ in res.py and run python res.py

Bias Tracing

Enter bias_tracing

Run the scripts bash scripts/gpt2.sh or bash scripts/roberta.sh.

Citing the paper

If this code or paper was useful, please consider using the following citation:

@article{xinxu2023EditBias,
    title={EditBias: Debiasing Language Models via Model Editing},
    author={Xin Xu, Wei Xu, Ningyu Zhang},
    year={2023},
    url={https://github.com/xxupiano/EditBias}
}