REMARK-LLM

[USENIX Scurity'24] Remark-llm: A robust and efficient watermarking framework for generative large language models

Paper

Environment

Important: Plz don't install huggingface transformer

pip install -r requirements.txt

Experiment

Train on the hc3 dataset

$ bash bash/run.sh train

Inference on the chatgpt dataset

$ bash bash/run.sh val

Citation

If you found our code/paper helpful, please kindly cite:

@inproceedings{zhang2024remark,
  title={$\{$REMARK-LLM$\}$: A Robust and Efficient Watermarking Framework for Generative Large Language Models},
  author={Zhang, Ruisi and Hussain, Shehzeen Samarah and Neekhara, Paarth and Koushanfar, Farinaz},
  booktitle={33rd USENIX Security Symposium (USENIX Security 24)},
  pages={1813--1830},
  year={2024}
}