[2301.10226] A Watermark for Large Language Models

arxiv.org · 6,513 words · saved by 1 readers

Potential harms of large language models can be mitigated by watermarking model output, i.e., embedding signals into generated text that are invisible to humans but algorithmically detectable from a short span of tokens. We propose a watermarking framework for proprietary language models. The watermark can be embedded with negligible impact on text quality, and can be detected using an efficient open-source algorithm without access to the language model API or parameters. The watermark works by selecting a randomized set of whitelist tokens before a word is generated, and then softly promoting use of whitelist tokens during sampling. We propose a statistical test for detecting the watermark with interpretable p-values, and derive an information-theoretic framework for analyzing the sensitivity of the watermark. We test the watermark using a multi-billion parameter model from the Open Pretrained Transformer (OPT) family, and discuss robustness and security.

A Watermark for Large Language Models John Kirchenbauer * Jonas Geiping * Yuxin Wen Jonathan Katz Ian Miers Tom Goldstein University of Maryland Abstract arXiv:2301.10226v4 [cs.LG] 1 May 2024 Potential harms of large language models can be mitigated Prompt…

[2301.10226] A Watermark for Large Language Models

related reading