{
  "id": "tokenization",
  "title": "Addressing Tokenization Inconsistency in Steganography and Watermarking Based on Large Language Models",
  "authors": [
    "Ruiyi Yan",
    "Yugo Murawaki"
  ],
  "author_parts": [
    {
      "given": "Ruiyi",
      "family": "Yan"
    },
    {
      "given": "Yugo",
      "family": "Murawaki"
    }
  ],
  "year": 2025,
  "venue_date": "2025-11",
  "venue": "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing",
  "venue_short": "EMNLP 2025",
  "publication_type": "conference",
  "publisher": "Association for Computational Linguistics",
  "pages": "7076–7098",
  "doi": "10.18653/v1/2025.emnlp-main.361",
  "arxiv_id": "2508.20718",
  "relation": "direct-extension",
  "categories": [
    "Tokenization",
    "Reliability",
    "Watermarking"
  ],
  "summary": "Studies tokenization inconsistencies between generated token sequences and the text a receiver retokenizes. It proposes stepwise verification for steganography and post-hoc rollback for watermark detection, evaluating how these changes affect reliability, text quality, and detection.",
  "relationship": "Section 4.1 explicitly uses Ziegler et al.'s arithmetic coding in all steganography experiments, extending the reliability layer around the 2019 approach rather than replacing its coding principle.",
  "limitation": "Steganographic extraction assumes an attacker does not modify the text. The authors report smaller watermarking gains because inconsistent tokens are infrequent and note that a complete theory of the inconsistency mechanism remains open.",
  "sources": [
    {
      "url": "https://aclanthology.org/2025.emnlp-main.361/",
      "evidence": "Full authors, November 2025 venue metadata, pages and DOI."
    },
    {
      "url": "https://arxiv.org/abs/2508.20718",
      "evidence": "First submission August 28, 2025."
    },
    {
      "url": "https://aclanthology.org/2025.emnlp-main.361.pdf",
      "evidence": "Page 1 links code; sections 2–4 describe methods; section 4.1 explicitly uses Ziegler et al. arithmetic coding; Limitations states threat model and scope of improvement."
    }
  ],
  "arxiv_url": "https://arxiv.org/abs/2508.20718",
  "earliest_date": "2025-08-28",
  "earliest_date_label": "First arXiv submission",
  "url": "https://aclanthology.org/2025.emnlp-main.361/",
  "pdf": "https://aclanthology.org/2025.emnlp-main.361.pdf",
  "code": "https://github.com/ryehr/Consistency",
  "category": "direct-extension",
  "contribution": "Studies tokenization inconsistencies between generated token sequences and the text a receiver retokenizes. It proposes stepwise verification for steganography and post-hoc rollback for watermark detection, evaluating how these changes affect reliability, text quality, and detection.",
  "canonical_url": "https://steganography.live/reading-list/papers/tokenization/"
}