Robust detection of watermarks for large language models under human edits

DOI10.1093/JRSSSB/QKAF056MaRDI QIDQ7240464zbMATH OpenFDO

Authors Feng Ruan, Huiyuan Wang, Qi Long, Weijie J. Su

Publication date 11 May 2026

Published in Journal of the Royal Statistical Society. Series B. Statistical Methodology (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1093/JRSSSB/QKAF056

zbMATH Keywords

adaptive optimalitydetection efficiencylarge language modelphase transitionrobust detectionwatermark

Mathematics Subject Classification ID

Statistics (62-XX)

Cites work

• A statistical framework of watermarks for large language models: pivot, detection efficiency and optimal rules
• A Statistical Problem Arising in the Theory of Detection of Signals in the Presence of Noise in a Multi-Channel System and Leading to Stable Distribution Laws
• Algorithmic High-Dimensional Robust Statistics
• Asymptotic Efficiency of Nonparametric Tests
• Covariate assisted screening and estimation
• Debiasing Watermarks for Large Language Models via Maximal Coupling
• Distribution-free tests for sparse heterogeneous mixtures
• Fast calculation of p-values for one-sided Kolmogorov-Smirnov type statistics
• Goodness-of-fit tests via phi-divergences
• Higher criticism for detecting sparse heterogeneous mixtures.
• Higher criticism for large-scale inference, especially for rare and weak effects
• Higher criticism: \(p\)-values and criticism
• scientific article; zbMATH DE number 1006115 (Why is no real title available?)
• Innovated higher criticism for detecting sparse signals in correlated noise
• Large deviations and Bahadur efficiency of the Khmaladze-Aki statistics
• Martingale Approach in the Theory of Goodness-of-Fit Tests
• Optimal adaptivity of signed-polygon statistics for network testing
• Optimal detection of heterogeneous and heteroscedastic mixtures
• Optimal Detection of Sparse Mixtures Against a Given Null Distribution
• Properties of higher criticism under strong dependence
• Pseudorandom error-correcting codes
• Rare and weak effects in large-scale inference: methods and phase diagrams
• Some problems of hypothesis testing leading to infinitely divisible distributions
• The intermediates take it all: asymptotics of higher criticism statistics and a powerful alternative based on equal local levels

This page was built for publication: Robust detection of watermarks for large language models under human edits

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7240464)