WaterMax: breaking the LLM watermark detectability-robustness-quality trade-off
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866916443756429312 |
|---|---|
| author | Giboulot, Eva Furon, Teddy |
| author_facet | Giboulot, Eva Furon, Teddy |
| contents | Watermarking is a technical means to dissuade malfeasant usage of Large Language Models. This paper proposes a novel watermarking scheme, so-called WaterMax, that enjoys high detectability while sustaining the quality of the generated text of the original LLM. Its new design leaves the LLM untouched (no modification of the weights, logits, temperature, or sampling technique). WaterMax balances robustness and complexity contrary to the watermarking techniques of the literature inherently provoking a trade-off between quality and robustness. Its performance is both theoretically proven and experimentally validated. It outperforms all the SotA techniques under the most complete benchmark suite. Code available at https://github.com/eva-giboulot/WaterMax. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_04808 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | WaterMax: breaking the LLM watermark detectability-robustness-quality trade-off Giboulot, Eva Furon, Teddy Cryptography and Security Computation and Language Machine Learning Watermarking is a technical means to dissuade malfeasant usage of Large Language Models. This paper proposes a novel watermarking scheme, so-called WaterMax, that enjoys high detectability while sustaining the quality of the generated text of the original LLM. Its new design leaves the LLM untouched (no modification of the weights, logits, temperature, or sampling technique). WaterMax balances robustness and complexity contrary to the watermarking techniques of the literature inherently provoking a trade-off between quality and robustness. Its performance is both theoretically proven and experimentally validated. It outperforms all the SotA techniques under the most complete benchmark suite. Code available at https://github.com/eva-giboulot/WaterMax. |
| title | WaterMax: breaking the LLM watermark detectability-robustness-quality trade-off |
| topic | Cryptography and Security Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2403.04808 |