University of Maryland
A public research university in the United States with research collaborations in fields including nuclear fusion and quantum computing.
Timeline (last 90 days)
The paper "A Watermark for Large Language Models" published by the University of Maryland in 2023 proposes dividing the vocabulary into green words and red words, with green words appearing at a significantly higher frequency in watermarked text than the statistical baseline.
The lexical distribution shift method was proposed and validated by research teams from institutions including Stanford University and the University of Maryland.
The KGW watermarking scheme was proposed by the University of Maryland in 2023 (Kirchenbauer et al.).
KGW算法由马里兰大学研究者Kirchenbauer等人于2023年在ICML上正式发表，是学术界引用量最高、影响力最大的LLM文本水印方案之一
KGW（Kirchenbauer-Geiping-Wen）算法由马里兰大学研究者于2023年提出，是学术界最具影响力的LLM文本水印方案之一
基于词表分割的文本水印方法由马里兰大学John Kirchenbauer等团队于2023年提出
All Facts (6)
The paper "A Watermark for Large Language Models" published by the University of Maryland in 2023 proposes dividing the vocabulary into green words and red words, with green words appearing at a significantly higher frequency in watermarked text than the statistical baseline.
50%UnverifiedThe lexical distribution shift method was proposed and validated by research teams from institutions including Stanford University and the University of Maryland.
50%UnverifiedThe KGW watermarking scheme was proposed by the University of Maryland in 2023 (Kirchenbauer et al.).
50%UnverifiedKGW算法由马里兰大学研究者Kirchenbauer等人于2023年在ICML上正式发表，是学术界引用量最高、影响力最大的LLM文本水印方案之一
50%UnverifiedKGW（Kirchenbauer-Geiping-Wen）算法由马里兰大学研究者于2023年提出，是学术界最具影响力的LLM文本水印方案之一
50%Unverified基于词表分割的文本水印方法由马里兰大学John Kirchenbauer等团队于2023年提出
50%