Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2505.01484 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918069552545792 |
|---|---|
| author | Abdalla, Pedro Vershynin, Roman |
| author_facet | Abdalla, Pedro Vershynin, Roman |
| contents | Given a text, can we determine whether it was generated by a large language model (LLM) or by a human? A widely studied approach to this problem is watermarking. We propose an undetectable and elementary watermarking scheme in the closed setting. Also, in the harder open setting, where the adversary has access to most of the model, we propose an unremovable watermarking scheme. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_01484 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | LLM Watermarking Using Mixtures and Statistical-to-Computational Gaps Abdalla, Pedro Vershynin, Roman Cryptography and Security Machine Learning Given a text, can we determine whether it was generated by a large language model (LLM) or by a human? A widely studied approach to this problem is watermarking. We propose an undetectable and elementary watermarking scheme in the closed setting. Also, in the harder open setting, where the adversary has access to most of the model, we propose an unremovable watermarking scheme. |
| title | LLM Watermarking Using Mixtures and Statistical-to-Computational Gaps |
| topic | Cryptography and Security Machine Learning |
| url | https://arxiv.org/abs/2505.01484 |