A Watermark for Large Language Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2023
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914779499593728 |
|---|---|
| author | Kirchenbauer, John Geiping, Jonas Wen, Yuxin Katz, Jonathan Miers, Ian Goldstein, Tom |
| author_facet | Kirchenbauer, John Geiping, Jonas Wen, Yuxin Katz, Jonathan Miers, Ian Goldstein, Tom |
| contents | 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 "green" tokens before a word is generated, and then softly promoting use of green 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2301_10226 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | A Watermark for Large Language Models Kirchenbauer, John Geiping, Jonas Wen, Yuxin Katz, Jonathan Miers, Ian Goldstein, Tom Machine Learning Computation and Language Cryptography and Security 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 "green" tokens before a word is generated, and then softly promoting use of green 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. |
| title | A Watermark for Large Language Models |
| topic | Machine Learning Computation and Language Cryptography and Security |
| url | https://arxiv.org/abs/2301.10226 |