mlsanityData to 7 Oct 2026

A Robust Semantics-based Watermark for Large Language Model against Paraphrasing

arXiv 2311.08721 ↗First posted 15 Nov 2023cs.CRcited by 71

On the map: LLM safety and security → Watermarking generated content and models

Abstract (arXiv)

Large language models (LLMs) have show great ability in various natural language tasks. However, there are concerns that LLMs are possible to be used improperly or even illegally. To prevent the malicious usage of LLMs, detecting LLM-generated text becomes crucial in the deployment of LLM applications. Watermarking is an effective strategy to detect the LLM-generated content by encoding a pre-defined secret watermark to facilitate the detection process. However, the majority of existing watermark methods leverage the simple hashes of precedent tokens to partition vocabulary. Such watermark can be easily eliminated by paraphrase and correspondingly the detection effectiveness will be greatly compromised. Thus, to enhance the robustness against paraphrase, we propose a semantics-based watermark framework SemaMark. It leverages the semantics as an alternative to simple hashes of tokens since the paraphrase will likely preserve the semantic meaning of the sentences. Comprehensive experiments are conducted to demonstrate the effectiveness and robustness of SemaMark under different paraphrases.

Key facts. Cited by 71 papers in the map (data through 7 Oct 2026). Placed in Watermarking generated content and models, one of 12 directions in LLM safety and security. First posted on arXiv 15 Nov 2023.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 3; 2024: 26; 2025: 20; 2026: 22.

71 papers.

Sort:

By area:

  1. 2023A Survey of Text Watermarking in the Era of Large Language Modelscited by 102
  2. 2024MarkLLM: An Open-Source Toolkit for LLM Watermarkingcited by 80
  3. 2024Watermark Stealing in Large Language Modelscited by 79
  4. 2023On the Learnability of Watermarks for Language Modelscited by 72
  5. 2024An Entropy-based Text Watermarking Detection Methodcited by 61
  6. 2024Adaptive Text Watermark for Large Language Modelscited by 60
  7. 2024No Free Lunch in LLM Watermarking: Trade-offs in Watermarking Design Choicescited by 47
  8. 2024Copyright Protection in Generative AI: A Technical Perspectivecited by 35
  9. 2024GumbelSoft: Diversified Language Model Watermarking via the GumbelMax-trickcited by 31
  10. 2024Token-Specific Watermarking with Enhanced Detectability and Semantic Coherence for Large Language Modelscited by 30
  11. 2024Deconstructing The Ethics of Large Language Models from Long-standing Issues to New-emerging Dilemmas: A Surveycited by 23
  12. 2024Black-Box Detection of Language Model Watermarkscited by 22
38 more
  1. 2024Double-I Watermark: Protecting Model Copyright for LLM Fine-tuningcited by 18
  2. 2024A Survey of AI-generated Text Forensic Systems: Detection, Attribution, and Characterizationcited by 18
  3. 2025BiMark: Unbiased Multilayer Watermarking for Large Language Modelscited by 18
  4. 2024Theoretically Grounded Framework for LLM Watermarking: A Distribution-Adaptive Approachcited by 17
  5. 2025Auditing Black-Box LLM APIs with a Rank-Based Uniformity Testcited by 16
  6. 2023Downstream Trade-offs of a Family of Text Watermarkscited by 14
  7. 2024A Watermark for Black-Box Language Modelscited by 14
  8. 2024Large Language Model Watermark Stealing With Mixed Integer Programmingcited by 13
  9. 2025VLA-Mark: A cross modal watermark for large vision-language alignment modelcited by 11
  10. 2025Defending LLM Watermarking Against Spoofing Attacks with Contrastive Representation Learningcited by 10
  11. 2024WaterJudge: Quality-Detection Trade-off when Watermarking Large Language Modelscited by 9
  12. 2025In-Context Watermarks for Large Language Modelscited by 9
  13. 2024Duwak: Dual Watermarks in Large Language Modelscited by 8
  14. 2025Enhancing LLM Watermark Resilience Against Both Scrubbing and Spoofing Attackscited by 8
  15. 2024Inevitable Trade-off between Watermark Strength and Speculative Sampling Efficiency for Language Modelscited by 7
  16. 2024Watermark under Fire: A Robustness Evaluation of LLM Watermarkingcited by 7
  17. 2025Position: LLM Watermarking Should Align Stakeholders' Incentives for Practical Adoptioncited by 7
  18. 2024The Impact of Prompts on Zero-Shot Detection of AI-Generated Textcited by 6
  19. 2024WaterPool: A Watermark Mitigating Trade-offs among Imperceptibility, Efficacy and Robustnesscited by 6
  20. 2025Diversity Boosts AI-Generated Text Detectioncited by 6
  21. 2024B4: A Black-Box Scrubbing Attack on LLM Watermarkscited by 5
  22. 2025From Trade-off to Synergy: A Versatile Symbiotic Watermarking Framework for Large Language Modelscited by 5
  23. 2024On Evaluating The Performance of Watermarked Machine-Generated Texts Under Adversarial Attackscited by 4
  24. 2024Discovering Spoofing Attempts on Language Model Watermarkscited by 4
  25. 2025Robustness Assessment and Enhancement of Text Watermarking for Google's SynthIDcited by 4
  26. 2025Paladin: Defending LLM-enabled Phishing Emails with a New Trigger-Tag Paradigmcited by 4
  27. 2025PRO: Enabling Precise and Robust Text Watermark for Open-Source LLMscited by 4
  28. 2024From Intentions to Techniques: A Comprehensive Taxonomy and Challenges in Text Watermarking for Large Language Modelscited by 3
  29. 2025BiMarker: Enhancing Text Watermark Detection for Large Language Models with Bipolar Watermarkscited by 3
  30. 2025ClusterMark: Towards Robust Watermarking for Autoregressive Image Generators with Visual Token Clusteringcited by 3
  31. 2026Distilling the Thought, Watermarking the Answer: A Principle Semantic Guided Watermark for Large Reasoning Modelscited by 3
  32. 2026A Unified Framework for LLM Watermarkscited by 3
  33. 2024Navigating the Shadows: Unveiling Effective Disturbances for Modern AI Content Detectorscited by 2
  34. 2025Sandcastles in the Storm: Revisiting the (Im)possibility of Strong Watermarkingcited by 2
  35. 2025SoK: Are Watermarks in LLMs Ready for Deployment?cited by 2
  36. 2025HeavyWater and SimplexWater: Distortion-Free LLM Watermarks for Low-Entropy Next-Token Predictionscited by 1
  37. 2025Can You Detect the Difference?cited by 1
  38. 2025Is Multilingual LLM Watermarking Truly Multilingual? Scaling Robustness to 100+ Languages via Back-Translationcited by 1

50 most cited of 71 All 71 (JSON)