It is shown that discovery is strictly harder than detection and that consistent classification is impossible across the parameter regime within this class of coordinatewise pivot-based localization rules, and an adaptive thresholding method is developed that attains the optimal discovery boundary and near-optimal disc...
José H. Blanchet, T. Cai, Xiang Li et al.· arXiv.org· 3 citations· ⚡1
Watermarking provides a principled way to authenticate text generated by large language models (LLMs). In practice, however, the final text may be mixed-source, with watermark evidence surviving at only a subset of token positions after rewriting, insertion, deletion, or paraphrasing. Although prior work has studied gl...
José H. Blanchet, T. Cai, Xiang Li et al.· 1 citation· ⚡1
This paper addresses the problem of optimally estimating the watermark proportion in mixed-source texts, and proposes efficient estimators for this class of methods, and derive minimax lower bounds for any measurable estimator based on pivotal statistics, showing that their estimators achieve these lower bounds.
Xiang Li, Garrett Wen, Weiqing He et al.· arXiv.org· 5 citations· ⚡1
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.