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
September 05, 2024
Chunting Zhou, Lili Yu, Arun Babu, Kushal Tirumala, Michihiro Yasunaga, Leonid Shamis, Jacob Kahn, Luke Zettlemoyer, Omer Levy, Xuezhe Ma
September 05, 2024
August 20, 2024
Ashish Shenoy, Yichao Lu, Srihari Jayakumar, Debojeet Chatterjee, Mohsen Moslehpour, Pierce Chuang, Abhay Harpale, Vikas Bhardwaj, Di Xu (SWE), Shicong Zhao, Ankit Ramchandani, Luna Dong, Anuj Kumar
August 20, 2024
August 15, 2024
Kamalika Chaudhuri, Chuan Guo, Laurens van der Maaten, Saeed Mahloujifar, Mark Tygert
August 15, 2024
August 11, 2024
Igor Tufanov, Karen Hambardzumyan, Javier Ferrando, Lena Voita
August 11, 2024
Foundational models
Latest news
Foundational models