June 7, 2024, 4:44 a.m. | Rohith Kuditipudi, John Thickstun, Tatsunori Hashimoto, Percy Liang

cs.LG updates on arXiv.org arxiv.org

arXiv:2307.15593v3 Announce Type: replace
Abstract: We propose a methodology for planting watermarks in text from an autoregressive language model that are robust to perturbations without changing the distribution over text up to a certain maximum generation budget. We generate watermarked text by mapping a sequence of random numbers -- which we compute using a randomized watermark key -- to a sample from the language model. To detect watermarked text, any party who knows the key can align the text to …

abstract arxiv autoregressive budget compute cs.cl cs.cr cs.lg distribution free generate language language model language models mapping maximum methodology numbers random replace robust text type watermarks

Senior Data Engineer

@ Displate | Warsaw

Solution Architect

@ Philips | Bothell - B2 - Bothell 22050

Senior Product Development Engineer - Datacenter Products

@ NVIDIA | US, CA, Santa Clara

Systems Engineer - 2nd Shift (Onsite)

@ RTX | PW715: Asheville Site W Asheville Greenfield Site TBD , Asheville, NC, 28803 USA

System Test Engineers (HW & SW)

@ Novanta | Barcelona, Spain

Senior Solutions Architect, Energy

@ NVIDIA | US, TX, Remote