Unverified50% confidenceFactExact time
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.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
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69% similarUnverifiedThe concept of AI text watermarking was first systematically proposed by John Kirchenbauer and others at the University of Maryland in their 2023 paper "A Watermark for Large Language Models".69% similarUnverifiedThe stronger the watermark signal, the higher the robustness, but the more restricted the vocabulary choices become, and the more easily text fluency and diversity decline.66% similarUnverifiedWatermarks in large language models are not visible messages, but a subtle statistical pattern introduced by the model when selecting tokens.65% similarUnverifiedA team from the University of Maryland proposed a lexical watermarking scheme that biases toward specific vocabulary during the token sampling process.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1004208API
curl https://kongchang.com/api/v1/knowledge/claims/1004208MCP
get_claim(id=1004208)