zjunlp / editbias

EditBias: Debiasing Stereotyped Language Models via Model Editing

EditBias: Debiasing Language Models via Model Editing

📃 Paper 💻 Code 🌏 Web

EditBias is an efficient model editing method to eliminate stereotyped bias from language models with small editor networks, including a debiasing loss to guide edits on partial parameters and a remaining loss to maintain the original language modeling abilities during editing. Experimental results show EditBias' excellent performance on debiasing and robustness of gender reverse and semantic generality.

🆕 News

• [Feb 2024] We released the paper and the refined code.
• [Dec 2023] Our idea was accepted by WiNLP 2023 and posted in EMNLP 2023!
• [Nov 2023] We released the code.

📌 Table of Contents

• 🛠️ Setup
• 💻 EditBias
  • ⌚️ Training Editor Networks
  • 🚀 Debiasing with Editor Networks
• 👀 Bias Tracing
• 📝 Citation
• ✨ Acknowledgements

🛠️ Setup

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

With StereoSet, editor networks are trained to modify partial parameters for debiasing at first. Then, the trained editor networks are used to conduct edits on language models and produce an unbiased model.

⌚️ Training Editor Networks

• Formatted datasets with train/dev/test (gender_test.json, race_test.json, religion_test.json) splits are in data/stereoset.
• Configurations are in config. Partial parameters to be edited are presented in model.
• Experimental scripts are in scripts. All hyper-parameters are in the scripts.
• For the ablation study on the remaining loss, set ifloc as False.
• Metrics can be found at the end of the training log.

For example, we use the following command to train the editor networks for GPT2-base:

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

• The parameters of the trained editor networks are stored in outputs/.../models/....bk. Record the path ending with .bk, like outputs/2024-02-08_18-51-18_4100072340/models/gpt2-.2024-02-08_18-51-18_4100072340.bk, as $p_1$.
• Metrics can be found at the end of the training log.

🚀 Debiasing with Editor Networks

• Set eval_only as True, archive as $p_1$, and val_set as the path of the test set file. The val_batch_size should be the same as the batch_size in training. See gpt2-base_val.sh for an example.
• Metrics can be found at the end of the debiasing log.
• For testing the robustness of gender reverse, set val_set as data/stereoset/gender_test_reverse.json.
• For testing the semantic generality, set val_set as data/stereoset/xxx_test_syn.json, where xxx is chosen from [gender, race, religion].

For example,

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

👀 Bias Tracing

Enter bias_tracing

📝 Citation

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

@article{xinxu24EditBias,
    title={EditBias: Debiasing Stereotyped Language Models via Model Editing},
    author={Xin Xu, Wei Xu, Ningyu Zhang},
    year={2024},
    url={https://github.com/zjunlp/EditBias}
}

✨ Acknowledgements

• Thanks for the original code from MEND.
• Thanks for StereoSet and all the baselines from bias-bench.
• For more model editing methods, please try EasyEdit.