TL;DR
This paper introduces a near-optimal, model-agnostic detection method for Gumbel watermarking, assuming i.i.d. next-token distributions, enhancing watermark detection reliability.
Contribution
It presents a simple, theoretically grounded detection mechanism that improves upon previous methods for Gumbel watermarking schemes.
Findings
The proposed detection method is near-optimal under certain assumptions.
It outperforms existing detection techniques in accuracy.
The mechanism is simple and model-agnostic.
Abstract
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.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.