Failures-of-Influence-Functions-in-LLMs

This is an official implementation of Do Influence Functions Work on Large Language Models?

Get Started

1. Run pip install -r requirements.txt
2. Run python finetune.py --dataset <dataset_name> --model <model_name> to train the model, and then evaluate the effectiveness of influence functions via python influence.py --lora <lora_name> --model <model_name>. We provide some examples of running a script of benchmarks in script.md.

You can change any hyperparameter if necessary. See finetune.py and influence.py for more details about the settings.

Citation

If you find this repo useful, please cite our paper.

@article{li2024influence,
  title={Do Influence Functions Work on Large Language Models?},
  author={Li, Zhe and Zhao, Wei and Li, Yige and Sun, Jun},
  journal={arXiv preprint arXiv:2409.19998},
  year={2024}
}

Contact

If you have any questions or want to discuss some details, please contact zheli@smu.edu.sg.

Acknowledgement

We appreciate the following github repos a lot for their valuable code base or datasets:

https://github.com/ykwon0407/DataInf

https://github.com/huggingface/peft