all InfoSec news
WaterMax: breaking the LLM watermark detectability-robustness-quality trade-off
March 11, 2024, 4:10 a.m. | Eva Giboulot, Furon Teddy
cs.CR updates on arXiv.org arxiv.org
Abstract: 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 …
arxiv breaking called cs.cl cs.cr cs.lg design generated high language language models large llm modification novel quality robustness technical text trade watermarking
More from arxiv.org / cs.CR updates on arXiv.org
IDEA: Invariant Defense for Graph Adversarial Robustness
1 day, 17 hours ago |
arxiv.org
FairCMS: Cloud Media Sharing with Fair Copyright Protection
1 day, 17 hours ago |
arxiv.org
Efficient unitary designs and pseudorandom unitaries from permutations
1 day, 17 hours ago |
arxiv.org
Jobs in InfoSec / Cybersecurity
SOC 2 Manager, Audit and Certification
@ Deloitte | US and CA Multiple Locations
Senior Security Researcher, SIEM
@ Huntress | Remote Canada
Senior Application Security Engineer
@ Revinate | San Francisco Bay Area
Cyber Security Manager
@ American Express Global Business Travel | United States - New York - Virtual Location
Incident Responder Intern
@ Bentley Systems | Remote, PA, US
SC2024-003533 Senior Online Vulnerability Assessment Analyst (CTS) - THU 9 May
@ EMW, Inc. | Mons, Wallonia, Belgium