KGW Watermark
An AI text watermarking scheme proposed in 2023 by Kirchenbauer et al. at the University of Maryland. It uses a context-seeded pseudorandom function to dynamically partition vocabulary into green and red lists, biasing model token selection toward green-list words to produce statistically detectable text watermarks.
Timeline (last 90 days)
The KGW watermarking scheme was proposed by the University of Maryland in 2023 (Kirchenbauer et al.).
The KGW watermark uses the preceding word or longer context as a seed to dynamically generate green/red list partitions through a pseudorandom function, causing the list to slide with the context.
All Facts (2)
The KGW watermarking scheme was proposed by the University of Maryland in 2023 (Kirchenbauer et al.).
50%UnverifiedThe KGW watermark uses the preceding word or longer context as a seed to dynamically generate green/red list partitions through a pseudorandom function, causing the list to slide with the context.
50%