Answers are grounded in this paper and cite their sources.
We present a publicly-detectable watermarking scheme for LMs: the detection algorithm contains no secret information, and it is executable by anyone. We embed a publicly-verifiable cryptographic signature into LM output using rejection sampling and prove that this produces unforgeable and distortion-free (i.e., undetectable without access to the public key) text output. We make use of error-correction to overcome periods of low entropy, a barrier for all prior watermarking schemes. We implement our scheme and find that our formal claims are met in practice.
No takes yet. Share an insight, caveat, or question.
Fairoze et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: