Unverified50% confidenceFactExact time
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.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedKirchenbauer等人的水印算法在生成每个token时，根据前文的哈希值将词汇表划分为绿名单和红名单，并轻微提升绿名单token被选中的概率73% similarUnverified统计学水印的核心原理是在大语言模型生成文本时，通过伪随机函数将候选token分为绿色列表和红色列表，略微提高绿色列表token的采样概率，通过z-test等统计检验方法检测62% similarUnverifiedKirchenbauer方案在模型原始logits上为绿名单token加上一个通常取1-2之间的固定偏置δ，再进行softmax归一化后采样53% similarUnverifiedThe generation pipeline of Diffusion-LM consists of three steps: mapping words to continuous vectors using learnable word embeddings, iteratively denoising via the diffusion model to obtain a sequence of continuous latent variables, and restoring continuous vectors to discrete text through a Rounding operation.53% similarUnverifiedPheno4D构成一个时序点云实例分割基准，植物这类非刚性、持续形变对象对模型的时序一致性和泛化能力提出更高要求52% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/900282API
curl https://kongchang.com/api/v1/knowledge/claims/900282MCP
get_claim(id=900282)