2024-06-30-christ24a.md

title	section	abstract	layout	series	publisher	issn	id	month	tex_title	firstpage	lastpage	page	order	cycles	bibtex_author	author	date	address	container-title	volume	genre	issued	pdf	extras
Undetectable Watermarks for Language Models	Original Papers	Recent advances in the capabilities of large language models such as GPT-4 have spurred increasing concern about our ability to detect AI-generated text. Prior works have suggested methods of embedding watermarks in model outputs, by *noticeably* altering the output distribution. We ask: Is it possible to introduce a watermark without incurring *any detectable* change to the output distribution? To this end, we introduce a cryptographically-inspired notion of undetectable watermarks for language models. That is, watermarks can be detected only with the knowledge of a secret key; without the secret key, it is computationally intractable to distinguish watermarked outputs from those of the original model. In particular, it is impossible for a user to observe any degradation in the quality of the text. Crucially, watermarks remain undetectable even when the user is allowed to adaptively query the model with arbitrarily chosen prompts. We construct undetectable watermarks based on the existence of one-way functions, a standard assumption in cryptography.	inproceedings	Proceedings of Machine Learning Research	PMLR	2640-3498	christ24a	0	Undetectable Watermarks for Language Models	1125	1139	1125-1139	1125	false	Christ, Miranda and Gunn, Sam and Zamir, Or	given family Miranda Christ given family Sam Gunn given family Or Zamir	2024-06-30		Proceedings of Thirty Seventh Conference on Learning Theory	247	inproceedings	date-parts 2024 6 30	https://proceedings.mlr.press/v247/christ24a/christ24a.pdf