TextSeal2：无失真 LLM 水印方案，兼容投机解码
TextSeal2: Distortion-Free LLM Watermarking with Copula Mixing and Memoryless Races
这篇论文解决了一个实际问题：水印遇到投机解码就失效，TextSeal2 让草稿 token 也带上同样的密钥，检测率还比前作高出 8-22 个百分点。
TextSeal2 是一个面向大语言模型的无失真水印方法，核心改动是用 Copula 将两个或多个密钥合并为单一虚拟密钥，检测器不再因不知密钥而平均分数。在投机解码场景下，被接受草稿的 token 用额外一次无记忆指数竞争验证，接受率保持不变，输出分布与目标模型完全一致。评测覆盖三个 LLM，在相同输出多样性下，TextSeal2 在 10^-3 假阳性率上比 TextSeal 多检出 8 到 22 个百分点的 100-token 回答，与 textGrain 相当且无需后处理 logits。在投机解码下，它对 200-token 回答的检出率分别比 TextSeal 和 textGrain 高 18 和 22 个百分点。
TextSeal2: Distortion-Free LLM Watermarking with Copula Mixing and Memoryless Races
We present TextSeal2, a distortion-free watermark for large language models that keeps its signal under speculative decoding: tokens that replace rejected drafts remain watermarked with the same keys, and the acceptance rate is unchanged. It also preserves output diversity and adds no measurable inference latency. Its predecessor, TextSeal, restores diversity in the Gumbel-max scheme by routing each token to one of two keys via a coin flip. This is suboptimal because the detector averages the scores from both keys without knowing which key was used. TextSeal2 instead merges two or more keys into a single virtual key and couples it to fresh noise through AutoMark's Gaussian copula. Any number of keys, including publicly shared keys, can be merged this way without diluting detection. Under speculative decoding, TextSeal2 verifies keyed drafts with one more exponential race, run on the leftovers of the draft races. By memorylessness, this race reproduces block verification, so the output follows the target model exactly. We evaluate TextSeal2 on three LLMs. At equal output token diversity, TextSeal2 detects 8 to 22 percentage points more 100-token answers than TextSeal at a false-positive rate of $10^{-3}$, and is comparable to textGrain while requiring no post-logits optimization. Under speculative decoding, it detects 18 and 22 percentage points more 200-token answers than TextSeal and textGrain, respectively, at the same acceptance rate.