Pith. sign in

REVIEW 51 references

Invisible Entropy: Towards Safe and Efficient Low-Entropy LLM Watermarking

T0 review · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A lightweight entropy classifier plus an adaptive threshold method can watermark and detect low-entropy LLM code outputs without querying the original model, matching much larger detectors at 99% fewer detection-phase parameters.

arxiv 2505.14112 v1 pith:IBJU3X6Q submitted 2025-05-20 cs.CL cs.CR

Tianle Gu , Zongqi Wang , Kexin Huang , Yuanqi Yao , Xiangliang Zhang , Yujiu Yang , Xiuying Chen This is my paper · ORCID
classification cs.CLcs.CR
keywords entropygreentokenwatermarkinglow-entropyoriginaladdressdatasets
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Watermarking tags AI-generated text by secretly favoring certain green words during generation, so the text can be recognized later. The standard KGW method works when the next word is hard to predict, but fails on code where tokens like "import numpy as" are almost forced. SWEET and EWD fix this by watermarking only high-entropy tokens, but the detector must call the original 15-billion-parameter language model to compute entropy, which is slow, expensive, and risks model leakage. Invisible Entropy replaces that call with a 130-million-parameter network: a tokenizer translator feeds the prefix into SimCSE embeddings, and a small MLP tagger predicts whether the next token will have high or low entropy. A Threshold Navigator then tunes the entropy threshold per sample, aiming to keep the watermark detectable while preserving natural code. The authors train the tagger on MBPP code using StarCoder's entropy as labels and test on HumanEval and MBPP, reporting detection accuracy close to methods that use the full model with 99% fewer detection-phase parameters. Main caveats are that reported numbers come from selecting the best configuration on the test set, the theoretical argument for the threshold rule is mathematically shaky, and the tagger's out-of-domain accuracy on HumanEval is only about 66-76%.
Extended reading notes

Core claim

From the paper's contributions: "Our proposed watermarking framework, IE, which integrates the three components, achieves a 99% reduction in parameter usage while delivering state-of-the-art detection performance." Concretely, a 130M-parameter entropy tagger plus adaptive threshold can replace a 15.5B-parameter LLM for entropy-based detection of watermarked code while keeping AUROC/TPR close to SWEET and EWD (HumanEval AUROC 0.941 vs 0.944; MBPP AUROC 0.892 vs 0.901).

Load-bearing premise

The entropy tagger, trained on MBPP train with StarCoder labels, must generalize to other code distributions and tokenizers well enough to keep detection accuracy high. Appendix C.3 Table 4 shows out-of-domain accuracy on HumanEval is only 66.61% at threshold 0.6 and 75.71% at 1.5; if this accuracy drops further on new code styles, the watermark becomes undetectable. The claim also depends on the last-token SimCSE embedding being a sufficient representation of the prefix for entropy prediction (Section 4.1).

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The method rests on five hand-set hyperparameters, several domain assumptions about embedding sufficiency and binary entropy classification, and one false independence assumption in the theoretical proof. No new physical entities are introduced.

free parameters (5)
  • green list proportion gamma = 0.5 (HumanEval), 0.25 (MBPP) for IE (Table 7)
    Chosen per dataset rather than derived; controls watermark strength and affects every reported detection metric.
  • bias delta = 3.0 for IE on both datasets (Table 7)
    Hand-set watermark strength; higher delta improves detectability but degrades code quality.
  • threshold search start tau0 = 1.5
    Initial high entropy threshold for the Threshold Navigator in Algorithm 4; hand-chosen without a stated selection rule.
  • threshold search granularity Delta = 0.3
    Step size for the Navigator; the authors acknowledge in the Limitations section that this fixed granularity limits flexibility.
  • Pass@1 drop allowance = 20%
    Selection rule in Appendix D: the paper reports the configuration with the highest UES among those lowering Pass@1 by at most 20%, a post-hoc selection criterion on the test set.
assumptions (6)
  • standard math KGW watermark generation and detection with green/red lists and the z-statistic is a valid background framework.
    Borrowed from Kirchenbauer et al. 2023 and used throughout as the base method for generation and detection.
  • standard math The number of green tokens in human text follows a normal distribution N(gamma times |T~|, gamma(1-gamma) times |T~|).
    Used in Section 6.2 for the Type-I error analysis; relies on the central limit theorem and an independence assumption.
  • domain assumption The SimCSE embedding of the last token captures enough context to predict the entropy of the next token.
    Section 4.1 uses only the last token's embedding as the representation of the whole prefix; this is an empirical design choice without a proof of sufficiency.
  • domain assumption Binary classification of high versus low entropy is sufficient for selective watermarking.
    Core premise of the Entropy Tagger; the paper argues exact entropy is unnecessary, which is plausible but not proven.
  • ad hoc to paper The partial derivative analysis in Section 6.2 treats |S|G and WR as independently variable quantities.
    The proof computes partial z / partial WR while holding |S|G fixed, even though both |S|G and WR are functions of the number of watermarked tokens. This assumption is false but is load-bearing for the claimed theory.
  • domain assumption StarCoder's entropy values are treated as ground truth labels for training the Entropy Tagger.
    Appendix C.1 computes exact entropy using StarCoder and uses it as y in the training set; if this reference entropy is biased, the tagger inherits the bias.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Invisible Entropy: Towards Safe and Efficient Low-Entropy LLM Watermarking." pith.science (2026). https://pith.science/paper/IBJU3X6Q

@misc{pith2026250514112,
  author       = {Pith},
  title        = {Pith review of: Invisible Entropy: Towards Safe and Efficient Low-Entropy LLM Watermarking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IBJU3X6Q}},
  note         = {Machine review of arXiv:2505.14112}
}
read the original abstract

Logit-based LLM watermarking traces and verifies AI-generated content by maintaining green and red token lists and increasing the likelihood of green tokens during generation. However, it fails in low-entropy scenarios, where predictable outputs make green token selection difficult without disrupting natural text flow. Existing approaches address this by assuming access to the original LLM to calculate entropy and selectively watermark high-entropy tokens. However, these methods face two major challenges: (1) high computational costs and detection delays due to reliance on the original LLM, and (2) potential risks of model leakage. To address these limitations, we propose Invisible Entropy (IE), a watermarking paradigm designed to enhance both safety and efficiency. Instead of relying on the original LLM, IE introduces a lightweight feature extractor and an entropy tagger to predict whether the entropy of the next token is high or low. Furthermore, based on theoretical analysis, we develop a threshold navigator that adaptively sets entropy thresholds. It identifies a threshold where the watermark ratio decreases as the green token count increases, enhancing the naturalness of the watermarked text and improving detection robustness. Experiments on HumanEval and MBPP datasets demonstrate that IE reduces parameter size by 99\% while achieving performance on par with state-of-the-art methods. Our work introduces a safe and efficient paradigm for low-entropy watermarking. https://github.com/Carol-gutianle/IE https://huggingface.co/datasets/Carol0110/IE-Tagger

Figures

Figures reproduced from arXiv: 2505.14112 by the authors.

Figure 1
Figure 1. Existing watermarking methods in low￾entropy scenarios face safety and cost challenges, while our method addresses them efficiently and securely. be applied at various stages, including logits gen￾eration (Kirchenbauer et al., 2023), token sam￾pling (Christ et al., 2024), and training (Sun et al., 2022, 2023; Gu et al., 2024). Logit-based water￾marking is cost-efficient, modifying probabilities before token selectio… view at source ↗
Figure 2
Figure 2. Overview of IE (Invisible Entropy). The model includes three components: the Unified Feature Extractor for tokenizer compatibility and feature extraction, the Entropy Tagger to predict if the next token’s entropy exceeds threshold τ , and the Threshold Navigator to optimize τ for effective watermarking, naturalness, and robustness. Tokens are color-coded as red (red list), green (green list), and gray (unwatermarked… view at source ↗
Figure 3
Figure 3. Analysis of the Entropy Tagger. (a) Comparison of applying the Entropy Tagger at different stages: generation-detection versus detection-only. (b) The relationship between Entropy Tagger accuracy and its effec￾tiveness in watermarking. (c) Demonstration of the superior performance of the Entropy Tagger compared to a surrogate model and randomly set entropy. higher-entropy tokens during detection. To explore the perf… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: (a) Type-I Error probability and the distribution of detection statistic [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Effectiveness of the Threshold Navigator. (a) Improved detectability and quality with the Navigator across δ. (b) Improved UES with the Navigator. (c) Generalizability to SWEET: Pass@1 vs. AUROC, demonstrating similar improvements. (d) UES comparison for SWEET, showing…
Figure 6
Figure 6. Figure 6: Results for various entropy thresholds. 12 [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Top-K tokens most frequently classified as low entropy tokens. [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 23 canonical work pages

  1. [1]

    Atallah, Victor Raskin, Michael Crogan, Christian Hempelmann, Florian Kerschbaum, Dina Mohamed, and Sanket Naik

    Mikhail J. Atallah, Victor Raskin, Michael Crogan, Christian Hempelmann, Florian Kerschbaum, Dina Mohamed, and Sanket Naik. 2001. Natural language watermarking: Design, analysis, and a proof-of-concept implementation. In Information Hiding, pages 185--200, Berlin, Heidelberg. Springer Berlin Heidelberg

    work page 2001

  2. [2]

    Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, et al. 2021. Program synthesis with large language models. arXiv preprint arXiv:2108.07732

    arXiv 2021

  3. [3]

    Brassil, S

    J.T. Brassil, S. Low, N.F. Maxemchuk, and L. O'Gorman. 1995. https://doi.org/10.1109/49.464718 Electronic marking and identification techniques to discourage document copying . IEEE Journal on Selected Areas in Communications, 13(8):1495--1504

    work page doi:10.1109/49.464718 1995

  4. [4]

    Liang Chen, Yatao Bian, Yang Deng, Deng Cai, Shuaiyi Li, Peilin Zhao, and Kam-Fai Wong. 2024. https://doi.org/10.18653/v1/2024.acl-long.496 W at ME : Towards lossless watermarking through lexical redundancy . In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 9166--9180, Bangkok, Thail...

    doi:10.18653/v1/2024.acl-long.496 2024

  5. [5]

    Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, Alex Ray, Raul Puri, Gretchen Krueger, Michael Petrov, Heidy Khlaaf, Girish Sastry, Pamela Mishkin, Brooke Chan, Scott Gray, Nick Ryder, Mikhail Pavlov, Alethea Power, Lukasz Kaiser, Mohammad Bavarian...

    arXiv 2021

  6. [6]

    Miranda Christ, Sam Gunn, and Or Zamir. 2024. Undetectable watermarks for language models. In The Thirty Seventh Annual Conference on Learning Theory, pages 1125--1139. PMLR

    work page 2024

  7. [7]

    Alexandre Duc, Stefan Dziembowski, and Sebastian Faust. 2014. Unifying leakage models: from probing attacks to noisy leakage. In Advances in Cryptology--EUROCRYPT 2014: 33rd Annual International Conference on the Theory and Applications of Cryptographic Techniques, Copenhagen, Denmark, May 11-15, 2014. Proceedings 33, pages 423--440. Springer

    work page 2014

  8. [8]

    Pierre Fernandez, Antoine Chaffin, Karim Tit, Vivien Chappelier, and Teddy Furon. 2023. Three bricks to consolidate watermarks for large language models. In 2023 IEEE International Workshop on Information Forensics and Security (WIFS), pages 1--6. IEEE

    work page 2023

Show all 51 references
  1. [9]

    Yu Fu, Deyi Xiong, and Yue Dong. 2024. Watermarking conditional text generation for ai detection: Unveiling challenges and a semantic-aware watermark remedy. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 18003--18011

    2024

  2. [10]

    Lang Gao, Xiangliang Zhang, Preslav Nakov, and Xiuying Chen. 2025. Shaping the safety boundaries: Understanding and defending against jailbreaks in large language models. ACL

    2025

  3. [11]

    Tianyu Gao, Xingcheng Yao, and Danqi Chen. 2021. SimCSE : Simple contrastive learning of sentence embeddings. In Empirical Methods in Natural Language Processing (EMNLP)

    2021

  4. [12]

    Sebastian Gehrmann, Hendrik Strobelt, and Alexander Rush. 2019. https://doi.org/10.18653/v1/P19-3019 GLTR : Statistical detection and visualization of generated text . In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics: System Demonstrat...

    2019 doi

  5. [13]

    Chenchen Gu, Xiang Lisa Li, Percy Liang, and Tatsunori Hashimoto. 2024. https://openreview.net/forum?id=9k0krNzvlV On the learnability of watermarks for language models . In The Twelfth International Conference on Learning Representations

    2024

  6. [14]

    Batu Guan, Yao Wan, Zhangqian Bi, Zheng Wang, Hongyu Zhang, Pan Zhou, and Lichao Sun. 2024. https://doi.org/10.18653/v1/2024.findings-emnlp.541 C ode IP : A grammar-guided multi-bit watermark for large language models of code . In Findings of the Association for Computational ...

    2024 doi

  7. [15]

    Abe Hou, Jingyu Zhang, Tianxing He, Yichen Wang, Yung-Sung Chuang, Hongwei Wang, Lingfeng Shen, Benjamin Van Durme, Daniel Khashabi, and Yulia Tsvetkov. 2024 a . https://doi.org/10.18653/v1/2024.naacl-long.226 S em S tamp: A semantic watermark with paraphrastic robustness for ...

    2024 doi

  8. [16]

    Abe Hou, Jingyu Zhang, Yichen Wang, Daniel Khashabi, and Tianxing He. 2024 b . https://doi.org/10.18653/v1/2024.findings-acl.98 k- S em S tamp: A clustering-based semantic watermark for detection of machine-generated text . In Findings of the Association for Computational Ling...

    2024 doi

  9. [17]

    Zhengmian Hu, Lichang Chen, Xidong Wu, Yihan Wu, Hongyang Zhang, and Heng Huang. 2024. https://openreview.net/forum?id=uWVC5FVidc Unbiased watermark for large language models . In The Twelfth International Conference on Learning Representations

    2024

  10. [18]

    John Kirchenbauer, Jonas Geiping, Yuxin Wen, Jonathan Katz, Ian Miers, and Tom Goldstein. 2023. A watermark for large language models. In International Conference on Machine Learning, pages 17061--17084. PMLR

    2023

  11. [19]

    John Kirchenbauer, Jonas Geiping, Yuxin Wen, Manli Shu, Khalid Saifullah, Kezhi Kong, Kasun Fernando, Aniruddha Saha, Micah Goldblum, and Tom Goldstein. 2024. https://openreview.net/forum?id=DEJIDCmWOz On the reliability of watermarks for large language models . In The Twelfth...

    2024

  12. [20]

    Kalpesh Krishna, Yixiao Song, Marzena Karpinska, John Frederick Wieting, and Mohit Iyyer. 2023. https://openreview.net/forum?id=WbFhFvjjKj Paraphrasing evades detectors of AI -generated text, but retrieval is an effective defense . In Thirty-seventh Conference on Neural Inform...

    2023

  13. [21]

    Rohith Kuditipudi, John Thickstun, Tatsunori Hashimoto, and Percy Liang. 2024. https://openreview.net/forum?id=FpaCL1MO2C Robust distortion-free watermarks for language models . Transactions on Machine Learning Research

    2024

  14. [22]

    Taehyun Lee, Seokhee Hong, Jaewoo Ahn, Ilgee Hong, Hwaran Lee, Sangdoo Yun, Jamin Shin, and Gunhee Kim. 2024. https://doi.org/10.18653/v1/2024.acl-long.268 Who wrote this code? watermarking for code generation . In Proceedings of the 62nd Annual Meeting of the Association for ...

    2024 doi

  15. [23]

    Raymond Li, Loubna Ben Allal, Yangtian Zi, Niklas Muennighoff, Denis Kocetkov, Chenghao Mou, Marc Marone, Christopher Akiki, Jia Li, Jenny Chim, et al. 2023. Starcoder: may the source be with you! arXiv preprint arXiv:2305.06161

    2023 arXiv

  16. [24]

    Aiwei Liu, Leyi Pan, Yijian Lu, Jingjing Li, Xuming Hu, Xi Zhang, Lijie Wen, Irwin King, Hui Xiong, and Philip Yu. 2024. A survey of text watermarking in the era of large language models. ACM Computing Surveys, 57(2):1--36

    2024

  17. [25]

    Yepeng Liu and Yuheng Bu. 2024. https://openreview.net/forum?id=7emOSb5UfX Adaptive text watermark for large language models . In Forty-first International Conference on Machine Learning

    2024

  18. [26]

    I Loshchilov. 2017. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101

    2017 arXiv

  19. [27]

    Yijian Lu, Aiwei Liu, Dianzhi Yu, Jingjing Li, and Irwin King. 2024. https://doi.org/10.18653/v1/2024.acl-long.630 An entropy-based text watermarking detection method . In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long P...

    2024 doi

  20. [28]

    Sumru Özsoy , Tunga Güngör, and Emre Sevinç

    Hasan Mesut Meral, Bülent Sankur, A. Sumru Özsoy , Tunga Güngör, and Emre Sevinç. 2009. https://doi.org/10.1016/j.csl.2008.04.001 Natural language watermarking via morphosyntactic alterations . Computer Speech & Language, 23(1):107--125

    2009 doi

  21. [29]

    Eric Mitchell, Yoonho Lee, Alexander Khazatsky, Christopher D Manning, and Chelsea Finn. 2023. https://proceedings.mlr.press/v202/mitchell23a.html D etect GPT : Zero-shot machine-generated text detection using probability curvature . In Proceedings of the 40th International Co...

    2023

  22. [30]

    Travis Munyer, Abdullah Tanvir, Arjon Das, and Xin Zhong. 2024. https://arxiv.org/abs/2305.05773 Deeptextmark: A deep learning-driven text watermarking approach for identifying large language model generated text . Preprint, arXiv:2305.05773

    2024 arXiv

  23. [31]

    Ansong Ni, Srini Iyer, Dragomir Radev, Veselin Stoyanov, Wen-Tau Yih, Sida Wang, and Xi Victoria Lin. 2023. https://proceedings.mlr.press/v202/ni23b.html LEVER : Learning to verify language-to-code generation with execution . In Proceedings of the 40th International Conference...

    2023

  24. [32]

    Lip Yee Por, KokSheik Wong, and Kok Onn Chee. 2012. https://doi.org/10.1016/j.jss.2011.12.023 Unispach: A text-based data hiding method using unicode space characters . Journal of Systems and Software, 85(5):1075--1082

    2012 doi

  25. [33]

    Stefano Giovanni Rizzo, Flavio Bertini, and Danilo Montesi. 2016. https://doi.org/10.1145/2938503.2938510 Content-preserving text watermarking through unicode homoglyph substitution . In Proceedings of the 20th International Database Engineering & Applications Symposium, IDEAS...

    2016

  26. [34]

    Ryoma Sato, Yuki Takezawa, Han Bao, Kenta Niwa, and Makoto Yamada. 2023. https://arxiv.org/abs/2310.08920 Embarrassingly simple text watermarks . Preprint, arXiv:2310.08920

    2023 arXiv

  27. [35]

    Irene Solaiman, Miles Brundage, Jack Clark, Amanda Askell, Ariel Herbert-Voss, Jeff Wu, Alec Radford, Gretchen Krueger, Jong Wook Kim, Sarah Kreps, Miles McCain, Alex Newhouse, Jason Blazakis, Kris McGuffie, and Jasmine Wang. 2019. https://arxiv.org/abs/1908.09203 Release stra...

    2019 arXiv

  28. [36]

    Congzheng Song and Ananth Raghunathan. 2020. Information leakage in embedding models. In Proceedings of the 2020 ACM SIGSAC conference on computer and communications security, pages 377--390

    2020

  29. [37]

    Zhensu Sun, Xiaoning Du, Fu Song, and Li Li. 2023. Codemark: Imperceptible watermarking for code datasets against neural code completion models. In Proceedings of the 31st ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineeri...

    2023

  30. [38]

    Zhensu Sun, Xiaoning Du, Fu Song, Mingze Ni, and Li Li. 2022. Coprotector: Protect open-source code against unauthorized training usage with data poisoning. In Proceedings of the ACM Web Conference 2022, pages 652--660

    2022

  31. [39]

    Mercan Topkara, Umut Topkara, and Mikhail J. Atallah. 2006. https://doi.org/10.1145/1178766.1178777 Words are not enough: sentence level natural language watermarking . In Proceedings of the 4th ACM International Workshop on Contents Protection and Security, MCPS '06, page 37–...

    2006

  32. [40]

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. 2023. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288

    2023 arXiv

  33. [41]

    Lean Wang, Wenkai Yang, Deli Chen, Hao Zhou, Yankai Lin, Fandong Meng, Jie Zhou, and Xu Sun. 2024. https://openreview.net/forum?id=JYu5Flqm9D Towards codable watermarking for injecting multi-bits information to LLM s . In The Twelfth International Conference on Learning Repres...

    2024

  34. [42]

    Zongqi Wang, Tianle Gu, Baoyuan Wu, and Yujiu Yang. 2025. https://arxiv.org/abs/2505.11541 Morphmark: Flexible adaptive watermarking for large language models . Preprint, arXiv:2505.11541

    2025 arXiv

  35. [43]

    Bram Wouters. 2024. https://openreview.net/forum?id=QGAeWRRe6e Optimizing watermarks for large language models . In Forty-first International Conference on Machine Learning

    2024

  36. [44]

    Yihan Wu, Zhengmian Hu, Hongyang Zhang, and Heng Huang. 2023. Dipmark: A stealthy, efficient and resilient watermark for large language models. arXiv preprint arXiv:2310.07710

    2023 arXiv

  37. [45]

    Hengyuan Xu, Liyao Xiang, Xingjun Ma, Borui Yang, and Baochun Li. 2024 a . Hufu: A modality-agnositc watermarking system for pre-trained transformers via permutation equivariance. arXiv preprint arXiv:2403.05842

    2024 arXiv

  38. [46]

    Xiaojun Xu, Yuanshun Yao, and Yang Liu. 2024 b . Learning to watermark llm-generated text via reinforcement learning. arXiv preprint arXiv:2403.10553

    2024 arXiv

  39. [47]

    Xi Yang, Kejiang Chen, Weiming Zhang, Chang Liu, Yuang Qi, Jie Zhang, Han Fang, and Nenghai Yu. 2023. https://arxiv.org/abs/2305.08883 Watermarking text generated by black-box language models . Preprint, arXiv:2305.08883

    2023 arXiv

  40. [48]

    Xi Yang, Jie Zhang, Kejiang Chen, Weiming Zhang, Zehua Ma, Feng Wang, and Nenghai Yu. 2022. https://doi.org/10.1609/aaai.v36i10.21415 Tracing text provenance via context-aware lexical substitution . Proceedings of the AAAI Conference on Artificial Intelligence, 36(10):11613--11621

    2022 doi

  41. [49]

    KiYoon Yoo, Wonhyuk Ahn, Jiho Jang, and Nojun Kwak. 2023. https://doi.org/10.18653/v1/2023.acl-long.117 Robust multi-bit natural language watermarking through invariant features . In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volum...

    2023 doi

  42. [50]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  43. [51]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.