Improving the Trade-off Between Watermark Strength and Speculative Sampling Efficiency for Language Models
Year: 2026
Type: conference-paper
Abstract: Watermarking is a principled approach for tracing the provenance of large language model (LLM) outputs, but its deployment in practice is hindered by inference inefficiency. Speculative sampling accelerates inference, with efficiency improving as the acceptance rate between draft and target models increases. Yet recent work reveals a fundamental trade-off: higher watermark strength reduces acceptance, preventing their simultaneous achievement. We revisit this trade-off and show it is not absolute. We introduce a quantitative measure of watermark strength that governs statistical detectability and is maximized when tokens are deterministic functions of pseudorandom numbers. Using this measure, we fully characterize the trade-off as a constrained optimization problem and derive explicit Pareto curves for two existing watermarking schemes. Finally, we introduce a principled mechanism that injects pseudorandomness into draft-token acceptance, ensuring maximal watermark strength while maintaining speculative sampling efficiency. Experiments further show that this approach improves detectability without sacrificing efficiency. Our findings uncover a principle that unites speculative sampling and watermarking, paving the way for their efficient and practical deployment.
Source: PubMed
Authors: He, Weiqing, Li, Xiang, Shen, Li, Su, Weijie, Long, Qi
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Awards: Robust privacy preserving distributed analysis platform for cancer research: addressing biased, incomplete data, Statistical Modeling of Alzheimer's Disease Progression Integrating Brain Imaging and -Omics Data, Advancing Neurosymbolic AI for Smart and Trustworthy Healthcare, Penn Artificial Intelligence and Technology Collaboratory for Healthy Aging, Robust and Interpretable Multi-modal AI/ML for Precision Medicine, AI Methods for Large Scale Epidemiological Studies using Patient Reports of Medication Adherence and Tolerability, Artificial Intelligence Strategies for Alzheimer's Disease Research
OpenAlex ID: W7127412983
FWCI:
Cites:
Cited by:
Related to: