Paper Detail

TextSeal: A Localized LLM Watermark for Provenance & Distillation Protection

Tom Sander, Hongyan Chang, Tomáš Souček, Tuan Tran, Valeriu Lacatusu, Sylvestre-Alvise Rebuffi, Alexandre Mourachko, Surya Parimi, Christophe Ropers, Rashel Moritz, Vanessa Stark, Hady Elsahar, Pierre Fernandez

arxiv Score 18.3

Published 2026-05-12 · First seen 2026-05-13

General AI

Abstract

We introduce TextSeal, a state-of-the-art watermark for large language models. Building on Gumbel-max sampling, TextSeal introduces dual-key generation to restore output diversity, along with entropy-weighted scoring and multi-region localization for improved detection. It supports serving optimizations such as speculative decoding and multi-token prediction, and does not add any inference overhead. TextSeal strictly dominates baselines like SynthID-text in detection strength and is robust to dilution, maintaining confident localized detection even in heavily mixed human/AI documents. The scheme is theoretically distortion-free, and evaluation across reasoning benchmarks confirms that it preserves downstream performance; while a multilingual human evaluation (6000 A/B comparisons, 5 languages) shows no perceptible quality difference. Beyond its use for provenance detection, TextSeal is also ``radioactive'': its watermark signal transfers through model distillation, enabling detection of unauthorized use.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
now
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@article{sander2026textseal,
  title = {TextSeal: A Localized LLM Watermark for Provenance \& Distillation Protection},
  author = {Tom Sander and Hongyan Chang and Tomáš Souček and Tuan Tran and Valeriu Lacatusu and Sylvestre-Alvise Rebuffi and Alexandre Mourachko and Surya Parimi and Christophe Ropers and Rashel Moritz and Vanessa Stark and Hady Elsahar and Pierre Fernandez},
  year = {2026},
  abstract = {We introduce TextSeal, a state-of-the-art watermark for large language models. Building on Gumbel-max sampling, TextSeal introduces dual-key generation to restore output diversity, along with entropy-weighted scoring and multi-region localization for improved detection. It supports serving optimizations such as speculative decoding and multi-token prediction, and does not add any inference overhead. TextSeal strictly dominates baselines like SynthID-text in detection strength and is robust to di},
  url = {https://arxiv.org/abs/2605.12456},
  keywords = {cs.CR, cs.CL, cs.LG},
  eprint = {2605.12456},
  archiveprefix = {arXiv},
}

Metadata

{}