Watermarking Makes Language Models Radioactive | Awesome LLM Papers Add your paper to Awesome LLM Papers

Watermarking Makes Language Models Radioactive

Tom Sander, Pierre Fernandez, Alain Durmus, Matthijs Douze, Teddy Furon . No Venue 2024

[Paper]   Search on Google Scholar   Search on Semantic Scholar
Compositional Generalization Fine Tuning Interdisciplinary Approaches Multimodal Semantic Representation Security Training Techniques

This paper investigates the radioactivity of LLM-generated texts, i.e. whether it is possible to detect that such input was used as training data. Conventional methods like membership inference can carry out this detection with some level of accuracy. We show that watermarked training data leaves traces easier to detect and much more reliable than membership inference. We link the contamination level to the watermark robustness, its proportion in the training set, and the fine-tuning process. We notably demonstrate that training on watermarked synthetic instructions can be detected with high confidence (p-value < 1e-5) even when as little as 5% of training text is watermarked. Thus, LLM watermarking, originally designed for detecting machine-generated text, gives the ability to easily identify if the outputs of a watermarked LLM were used to fine-tune another LLM.

Similar Work