Refined Detection for Gumbel Watermarking
arXiv:2603.3001767.4
AI Analysis
This work addresses watermarking detection for AI-generated text, but it is incremental as it builds on an existing scheme.
The authors tackled the problem of improving detection for the Gumbel watermarking scheme by proposing a new mechanism that is proven to be near-optimal among model-agnostic schemes under i.i.d. sampling assumptions.
We propose a simple detection mechanism for the Gumbel watermarking scheme proposed by Aaronson (2022). The new mechanism is proven to be near-optimal in a problem-dependent sense among all model-agnostic watermarking schemes under the assumption that the next-token distribution is sampled i.i.d.