Optimal Detection for Language Watermarks with Pseudorandom Collision

Text watermarking plays a crucial role in ensuring the traceability and accountability of large language model (LLM) outputs and mitigating misuse. While promising, most existing methods assume perfect pseudorandomness. In practice, repetition in generated text induces collisions that create structured dependence, compromising Type I error control and invalidating standard analyses. We introduce a statistical framework that captures this structure through a hierarchical two-layer partition. At its core is the concept of minimal units -- the smallest groups treatable as independent across units while permitting dependence within. Using minimal units, we define a non-asymptotic efficiency measure and cast watermark detection as a minimax hypothesis testing problem. Applied to Gumbel-max and inverse-transform watermarks, our framework produces closed-form optimal rules. It explains why discarding repeated statistics often improves performance and shows that within-unit dependence must be addressed unless degenerate. Both theory and experiments confirm improved detection power with rigorous Type I error control. These results provide the first principled foundation for watermark detection under imperfect pseudorandomness, offering both theoretical insight and practical guidance for reliable tracing of model outputs.

Key Contributions

• Introduces a hierarchical two-layer partition framework with 'minimal units' to model structured dependence caused by pseudorandom collisions in repeated text
• Casts watermark detection as a minimax hypothesis testing problem and derives closed-form optimal detection rules for Gumbel-max and inverse-transform watermarks
• Provides the first principled statistical foundation for watermark detection under imperfect pseudorandomness, with both theoretical guarantees and improved empirical detection power

🛡️ Threat Analysis

Details

Similar Papers