February 21, 2024
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 < 10−5) 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.
Publisher
ARXIV
April 14, 2024
Heng-Jui Chang, Ning Dong (AI), Ruslan Mavlyutov, Sravya Popuri, Andy Chung
April 14, 2024
December 20, 2023
François Charton, Kristin Lauter, Cathy Li, Mark Tygert
December 20, 2023
December 07, 2023
Hakan Inan, Kartikeya Upasani, Jianfeng Chi, Rashi Rungta, Krithika Iyer, Yuning Mao, Davide Testuggine, Madian Khabsa
December 07, 2023
December 06, 2023
Mattia Atzeni, Mike Plekhanov, Frederic Dreyer, Nora Kassner, Simone Merello, Louis Martin, Nicola Cancedda
December 06, 2023
Product experiences
Foundational models
Product experiences
Latest news
Foundational models