OPTML-Group/GradDiff-WMDP

1---2license: mit3datasets:4- cais/wmdp5language:6- en7base_model:8- HuggingFaceH4/zephyr-7b-beta9pipeline_tag: text-generation10library_name: transformers11tags:12- unlearn13- machine-unlearning14- llm-unlearning15- data-privacy16- large-language-models17- trustworthy-ai18- trustworthy-machine-learning19- language-model20---21 22# GradDiff-Unlearned Model on Task "WMDP"23 24## Model Details25 26- **Unlearning**:27  - **Task**: [🤗datasets/cais/wmdp wmdp-bio](https://huggingface.co/datasets/cais/wmdp)28  - **Method**: GradDiff29  - **Smoothness Optimization**: None30- **Origin Model**: [🤗HuggingFaceH4/zephyr-7b-beta](https://huggingface.co/HuggingFaceH4/zephyr-7b-beta)31- **Code Base**: [github.com/OPTML-Group/Unlearn-Smooth](https://github.com/OPTML-Group/Unlearn-Smooth)32- **Research Paper**: ["Towards LLM Unlearning Resilient to Relearning Attacks: A Sharpness-Aware Minimization Perspective and Beyond"](https://arxiv.org/abs/2502.05374)33 34## Loading the Model35 36```python37import torch38from transformers import AutoModelForCausalLM39 40model = AutoModelForCausalLM.from_pretrained("OPTML-Group/GradDiff-WMDP", torch_dtype=torch.bfloat16, trust_remote_code=True)41```42 43## Citation44 45If you use this model in your research, please cite: 46```47@article{fan2025towards,48  title={Towards LLM Unlearning Resilient to Relearning Attacks: A Sharpness-Aware Minimization Perspective and Beyond},49  author={Fan, Chongyu and Jia, Jinghan and Zhang, Yihua and Ramakrishna, Anil and Hong, Mingyi and Liu, Sijia},50  journal={arXiv preprint arXiv:2502.05374},51  year={2025}52}53```54 55## Reporting Issues56 57Reporting issues with the model: [github.com/OPTML-Group/Unlearn-Smooth](https://github.com/OPTML-Group/Unlearn-Smooth)