Unverified50% confidenceFactExact time
KGW算法在每个token生成时利用前一个或前几个token作为伪随机种子将词表划分为绿名单和红名单，采样时对绿名单token施加正向偏置，检测时统计绿名单token比例是否显著高于随机基线
1
Sources
50%
Confidence
Long-term
Relevance
8/27/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe green list/red list scheme works by using the previous token as a hash seed at each token generation step to divide the vocabulary into a green list and a red list, applying a positive logit bias to green-list tokens, and detecting whether the proportion of green-list tokens is significantly above 50% during verification.78% similarUnverifiedWhen generating each token, the watermarking algorithm randomly partitions the vocabulary into a green list and a red list based on a hash of the preceding context, and slightly increases the probability of green list tokens being selected.77% similarUnverifiedKirchenbauer方案通常将前一个token的ID与保密密钥拼接后输入哈希函数（如SHA-256）生成伪随机数种子，前γ比例（通常γ=0.5）的token归入绿名单71% similarUnverifiedKirchenbauer方案根据前文哈希值将词汇表随机划分为绿名单和红名单，通常γ=0.5，并在绿名单token的logits上加固定偏置δ（通常取1-2之间）68% similarUnverifiedRandom Forest randomly samples a subset of features at each node split to search for the optimal split, generating diversity among trees to reduce variance.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/807802API
curl https://kongchang.com/api/v1/knowledge/claims/807802MCP
get_claim(id=807802)