Elamur Research⌕
☰
home / papers / 2026 / Verifiable LLM-Generated Text Detection via Projected Semantic-Structural Distributions

Verifiable LLM-Generated Text Detection via Projected Semantic-Structural Distributions

Paper · 2026 · Ruochong Xiong, Qien Li, Wangwang Lian et al.

Summary

The paper targets detection of machine-generated text (MGT). Current detectors rely on proxy-model probabilities or a single semantic signal, which brings distribution misalignment, high compute costs, and little transparency. The authors found that, within a projected space combining semantics and syntax, MGT centroids move in one consistent direction away from human-written text (HWT) across many local syntactic configurations.

Their framework, ProSSD, works in three steps:

  • Projection: RoBERTa embeddings are mapped into a compact, label-aware subspace using supervised subspace projection (SSP) rather than PCA.
  • Modeling: semantic distributions are modeled conditioned on syntactic structure.
  • Scoring: texts are scored with a modified Mahalanobis distance, derived from a likelihood ratio test and weighted by Wasserstein distance.

No task-specific fine-tuning is needed. On DetectRL, which includes models such as GPT-5.1, ProSSD performs strongly in cross-domain, cross-model, and adversarial settings. It is also fast, memory-light, and data-efficient. Because every decision is a verifiable statistical test, the method matters for forensics and content moderation.

Key points

  • Efficiency: about 0.08 s per sample and 1.35 GB of GPU memory, 74.62% less than the runner-up. Its reference AUROC is 99.97, compared with 99.59 for RepreGuard and 92.52 for Binoculars.
  • Ablation: replacing SSP with PCA lowers XSum AUC from 99.21 to 94.75. Dropping the human/machine contrast lowers Review AUC to 88.70.
  • Settings and data needs: the best N-gram window is k = 4. Performance plateaus after roughly 400 reference samples and remains strong on short texts of 20–80 tokens.
  • Interpretability findings: according to the authors, MGT shows a consistent, directional shift in its semantics as well as a collapse of its semantic variety. Formal domains rely on logical connectives as cues, while subjective texts rely on pronouns and adjectives.
  • Limitations and availability: heavily polished or hybrid human-machine text is not handled well, and source attribution is not supported. Code and data are at github.com/RuoChoXio/ProSSD.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Peking UniversityBeijing · China315 works · 43,268 citations🎓Central South UniversityChangsha · China23 works · 657 citations🏢GeelyHangzhou · China2 works

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

Cite

Cite as Ruochong Xiong, Qien Li, Wangwang Lian, Yulong Wan, Hanlin Xue, Zhouxing Tan, Han Yang, Fengyu Lu and Junfei Liu. 2026. Verifiable LLM-Generated Text Detection via Projected Semantic-Structural Distributions. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.638

@inproceedings{xiong2026verifiable,
  title = {{Verifiable LLM-Generated Text Detection via Projected Semantic-Structural Distributions}},
  author = {Ruochong Xiong, Qien Li, Wangwang Lian, Yulong Wan, Hanlin Xue, Zhouxing Tan, Han Yang, Fengyu Lu and Junfei Liu},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.638},
  url = {https://doi.org/10.18653/v1/2026.acl-long.638}
}

References

67 references, 24 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Alim Al Ayub Ahmed, Ayman Aljabouh, Praveen Kumar Donepudi, and Myung Suh Choi. 2021. Detecting fake news using machine learning : A systematic literature review. CoRR, abs/2102.04458. — [arXiv:2102.04458](https://arxiv.org/abs/2102.04458) find
  2. Theodore Wilbur Anderson, Theodore Wilbur Anderson, Theodore Wilbur Anderson, Theodore Wilbur Anderson, and Etats-Unis Mathématicien. 1958. An introduction to multivariate statistical analysis, volume 2. Wiley New York. find
  3. Rohan Anil, Andrew M. Dai, Orhan Firat, Melvin Johnson, Dmitry Lepikhin, Alexandre Passos, Siamak Shakeri, Emanuel Taropa, Paige Bailey, Zhifeng Chen, Eric Chu, Jonathan H. Clark, Laurent El Shafey, Yanping Huang, Kathy Meier-Hellstern, Gaurav Mishra, Erica Moreira, Mark Omernick, Kevin Robinson, an… — PaLM 2 Technical Report
  4. Anthropic. 2025. Claude 4. Anthropic Blog. find
  5. Guangsheng Bao, Yanbin Zhao, Zhiyang Teng, Linyi Yang, and Yue Zhang. 2024. Fast-detectgpt: Efficient zero-shot detection of machine-generated text via conditional probability curvature. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024.…
  6. Yinpeng Cai, Lexin Li, and Linjun Zhang. 2025. A statistical hypothesis testing framework for data misappropriation detection in large language models. CoRR, abs/2501.02441. — [arXiv:2501.02441](https://arxiv.org/abs/2501.02441) find
  7. Jiaqi Chen, Xiaoye Zhu, Tianyang Liu, Ying Chen, Xinhui Chen, Yiwen Yuan, Chak Tou Leong, Zuchao Li, Tang Long, Lei Zhang, Chenyu Yan, Guanghao Mei, Jie Zhang, and Lefei Zhang. 2025a. Imitate before detect: Aligning machine stylistic preference for machine-revised text detection. In AAAI-25, Sponsor… find
  8. Xin Chen, Junchao Wu, Shu Yang, Runzhe Zhan, Zeyu Wu, Ziyang Luo, Di Wang, Min Yang, Lidia S. Chao, and Derek F. Wong. 2025b. Repreguard: Detecting llm-generated text by revealing hidden representation patterns. CoRR, abs/2508.13152. find
  9. Miranda Christ, Sam Gunn, and Or Zamir. 2024. Undetectable watermarks for language models. In The Thirty Seventh Annual Conference on Learning Theory, June 30 - July 3, 2023, Edmonton, Canada, volume 247 of Proceedings of Machine Learning Research, pages 1125–1139. PMLR. find
  10. Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Unsupervised cross-lingual representation learning at scale. CoRR, abs/1911.02116.
  11. Harald Cramér and Herman Wold. 1936. Some theorems on distribution functions. Journal of the London Mathematical Society, 1(4):290–294. find
  12. Evan Crothers, Nathalie Japkowicz, and Herna L. Viktor. 2023. Machine-generated text: A comprehensive survey of threat models and detection methods. IEEE Access, 11:70977–71002. find
  13. Dorottya Demszky, Diyi Yang, David S Yeager, Christopher J Bryan, Margarett Clapper, Susannah Chandhok, Johannes C Eichstaedt, Cameron Hecht, Jeremy Jamieson, Meghann Johnson, and 1 others. 2023. Using large language models in psychology. Nature Reviews Psychology, 2(11):688–701. find
  14. Anil R Doshi and Oliver P Hauser. 2024. Generative ai enhances individual creativity but reduces the collective diversity of novel content. Science advances, 10(28):eadn5290. find
  15. Matthew Durward and Christopher Thomson. 2024. Evaluating vocabulary usage in llms. In Proceedings of the 19th Workshop on Innovative Use of NLP for Building Educational Applications, BEA 2024, Mexico City, Mexico, June 20, 2024, pages 266–282. Association for Computational Linguistics. find
  16. Tiziano Fagni, Fabrizio Falchi, Margherita Gambini, Antonio Martella, and Maurizio Tesconi. 2020. Tweepfake: about detecting deepfake tweets. CoRR, abs/2008.00036.
  17. Angela Fan, Mike Lewis, and Yann Dauphin. 2018. Hierarchical neural story generation. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 889–898, Melbourne, Australia. Association for Computational Linguistics.
  18. William Feller. 1991. An introduction to probability theory and its applications, Volume 2, volume 2. John Wiley & Sons. find
  19. Leon Fröhling and Arkaitz Zubiaga. 2021. Feature-based detection of automated language models: tackling gpt-2, GPT-3 and grover. PeerJ Comput. Sci., 7:e443. find
  20. Jiachen Fu, Chun-Le Guo, and Chongyi Li. 2025. Detectanyllm: Towards generalizable and robust detection of machine-generated text across domains and models. CoRR, abs/2509.14268. — [arXiv:2509.14268](https://arxiv.org/abs/2509.14268) find
  21. Zheng Ge, Songtao Liu, Zeming Li, Osamu Yoshie, and Jian Sun. 2021. OTA: optimal transport assignment for object detection. In IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2021, virtual, June 19-25, 2021, pages 303–312. Computer Vision Foundation / IEEE. find
  22. Sebastian Gehrmann, Hendrik Strobelt, and Alexander M. Rush. 2019. GLTR: statistical detection and visualization of generated text. In Proceedings of the 57th Conference of the Association for Computational Linguistics, ACL 2019, Florence, Italy, July 28 - August 2, 2019, Volume 3: System Demonstrat…
  23. Noah Golowich and Ankur Moitra. 2024. Edit distance robust watermarks via indexing pseudorandom codes. In Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver, BC, Canada, December 10 - 15, 2024. find
  24. Google DeepMind. 2025. Gemini 3 flash model card. Technical Report. find
  25. Abhimanyu Hans, Avi Schwarzschild, Valeriia Cherepanova, Hamid Kazemi, Aniruddha Saha, Micah Goldblum, Jonas Geiping, and Tom Goldstein. 2024. Spotting llms with binoculars: Zero-shot detection of machine-generated text. In Forty-first International Conference on Machine Learning, ICML 2024, Vienna,…
  26. Beizhe Hu, Qiang Sheng, Juan Cao, Yang Li, and Danding Wang. 2025. Llm-generated fake news induces truth decay in news ecosystem: A case study on neural news recommendation. In Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2025… find
  27. Xiaomeng Hu, Pin-Yu Chen, and Tsung-Yi Ho. 2023. RADAR: robust ai-text detection via adversarial learning. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023.
  28. Daphne Ippolito, Daniel Duckworth, Chris Callison-Burch, and Douglas Eck. 2020. Automatic detection of generated text is easiest when humans are fooled. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, ACL 2020, Online, July 5-10, 2020, pages 1808–1822. Ass…
  29. Lei Jiang, Desheng Wu, and Xiaolong Zheng. 2025. Sendetex: Sentence-level ai-generated text detection for human-ai hybrid content via style and context fusion. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 5287–5302. find
  30. Hoang Anh Just, Feiyang Kang, Tianhao Wang, Yi Zeng, Myeongseob Ko, Ming Jin, and Ruoxi Jia. 2023. LAVA: data valuation without pre-specified learning algorithms. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023. OpenReview.net. find
  31. 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, ICML 2023, 23-29 July 2023, Honolulu, Hawaii, USA, volume 202 of Proceedings of Machine Learning Research, pages 170…
  32. Ryuto Koike, Masahiro Kaneko, Ayana Niwa, Preslav Nakov, and Naoaki Okazaki. 2025. Exagpt: Example-based machine-generated text detection for human interpretability. CoRR, abs/2502.11336. find
  33. Ryuto Koike, Masahiro Kaneko, and Naoaki Okazaki. 2024. OUTFOX: llm-generated essay detection through in-context learning with adversarially generated examples. In Thirty-Eighth AAAI Conference on Artificial Intelligence, AAAI 2024, Thirty-Sixth Conference on Innovative Applications of Artificial In… find
  34. Kalpesh Krishna, Yixiao Song, Marzena Karpinska, John Wieting, and Mohit Iyyer. 2023. Paraphrasing evades detectors of ai-generated text, but retrieval is an effective defense. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, N…
  35. Richard J Larsen and Morris L Marx. 2005. An introduction to mathematical statistics, volume 106. Prentice Hall Hoboken, NJ. find
  36. Olivier Ledoit and Michael Wolf. 2004. A well-conditioned estimator for large-dimensional covariance matrices. Journal of multivariate analysis, 88(2):365–411. find
  37. Donghyeok Lee, Christina Todorova, and Alireza Dehghani. 2024. Ethical risks and future direction in building trust for large language models application under the EU AI act. In Proceedings of the 2024 Conference on Human Centred Artificial Intelligence - Education and Practice, HCAIep 2024, Naples,… find
  38. Xiaoze Liu, Ting Sun, Tianyang Xu, Feijie Wu, Cunxiang Wang, Xiaoqian Wang, and Jing Gao. 2024. SHIELD: Evaluation and defense strategies for copyright compliance in LLM text generation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 1640–1670, Miami… find
  39. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized BERT pretraining approach. CoRR, abs/1907.11692.
  40. Eric Mitchell, Yoonho Lee, Alexander Khazatsky, Christopher D. Manning, and Chelsea Finn. 2023. Detectgpt: Zero-shot machine-generated text detection using probability curvature. In International Conference on Machine Learning, ICML 2023, 23-29 July 2023, Honolulu, Hawaii, USA, volume 202 of Proceed…
  41. Ryo Nagata, Hiroya Takamura, Naoki Otani, and Yoshifumi Kawasaki. 2023. Variance matters: Detecting semantic differences without corpus/word alignment. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023, pages 15609–… find
  42. Shashi Narayan, Shay B. Cohen, and Mirella Lapata. 2018. Don’t give me the details, just the summary! topic-aware convolutional neural networks for extreme summarization. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 1797–1807, Brussels, Belgium. As…
  43. OpenAI. 2023. Introducing chatgpt. OpenAI Blog. find
  44. OpenAI. 2025. Gpt-5.1. OpenAI Blog. find
  45. Hyeonchu Park, Byungjun Kim, and Bugeun Kim. 2025. DART: an AIGT detector using AMR of rephrased text. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL 2025 - Volume 2: Short Papers, Albu… find
  46. Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D. Manning, Stefano Ermon, and Chelsea Finn. 2023. Direct preference optimization: Your language model is secretly a reward model. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Sy…
  47. Vinu Sankar Sadasivan, Aounon Kumar, Sriram Balasubramanian, Wenxiao Wang, and Soheil Feizi. 2025. Can ai-generated text be reliably detected? stress testing AI text detectors under various attacks.
  48. Trans. Mach. Learn. Res., 2025. find
  49. Areg Mikael Sarvazyan, José Ángel González, Paolo Rosso, and Marc Franco-Salvador. 2023. Supervised machine-generated text detectors: Family and scale matters. In Experimental IR Meets Multilinguality, Multimodality, and Interaction - 14th International Conference of the CLEF Association, CLEF 2023,… find
  50. John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. 2017. Proximal policy optimization algorithms. CoRR, abs/1707.06347. find
  51. Irene Solaiman, Miles Brundage, Jack Clark, Amanda Askell, Ariel Herbert-Voss, Jeff Wu, Alec Radford, and Jasmine Wang. 2019. Release strategies and the social impacts of language models. CoRR, abs/1908.09203.
  52. Jinyan Su, Terry Yue Zhuo, Di Wang, and Preslav Nakov. 2023. Detectllm: Leveraging log rank information for zero-shot detection of machine-generated text. In Findings of the Association for Computational Linguistics: EMNLP 2023, Singapore, December 6-10, 2023, pages 12395–12412. Association for Comp…
  53. Chenxia Tang, Jianchun Liu, Hongli Xu, and Liusheng Huang. 2025. Top-nσ: Eliminating noise in logit space for robust token sampling of LLM. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 10758–10774, Vienna, Austria. Associat… find
  54. Qwen Team. 2024. Qwen2.5: A party of foundation models. find
  55. Eduard Tulchinskii, Kristian Kuznetsov, Laida Kushnareva, Daniil Cherniavskii, Sergey I. Nikolenko, Evgeny Burnaev, Serguei Barannikov, and Irina Piontkovskaya. 2023. Intrinsic dimension estimation for robust detection of ai-generated texts. In Advances in Neural Information Processing Systems 36: A…
  56. Cédric Villani and 1 others. 2008. Optimal transport: old and new, volume 338. Springer. find
  57. Yuxia Wang, Jonibek Mansurov, Petar Ivanov, Jinyan Su, Artem Shelmanov, Akim Tsvigun, Osama Mohammed Afzal, Tarek Mahmoud, Giovanni Puccetti, Thomas Arnold, Alham Fikri Aji, Nizar Habash, Iryna Gurevych, and Preslav Nakov. 2024. M4gtbench: Evaluation benchmark for black-box machine-generated text de… find
  58. Herman Wold. 1966. Estimation of principal components and related models by iterative least squares. Multivariate analysis, pages 391–420. find
  59. Junchao Wu, Shu Yang, Runzhe Zhan, Yulin Yuan, Lidia Sam Chao, and Derek Fai Wong. 2025a. A survey on LLM-generated text detection: Necessity, methods, and future directions. Computational Linguistics, 51(1):275–338.
  60. Junchao Wu, Runzhe Zhan, Derek F. Wong, Shu Yang, Xuebo Liu, Lidia S. Chao, and Min Zhang. 2025b. Who wrote this? the key to zero-shot llm-generated text detection is gecscore. In Proceedings of the 31st International Conference on Computational Linguistics, COLING 2025, Abu Dhabi, UAE, January 19-2…
  61. Junchao Wu, Runzhe Zhan, Derek F. Wong, Shu Yang, Xinyi Yang, Yulin Yuan, and Lidia S. Chao. 2024. find
  62. Detectrl: Benchmarking llm-generated text detection in real-world scenarios. In Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver, BC, Canada, December 10 - 15, 2024. xAI. 2025. Grok-4.1 model card. Technica… find
  63. Xianjun Yang, Wei Cheng, Yue Wu, Linda Ruth Petzold, William Yang Wang, and Haifeng Chen. 2024. DNA-GPT: divergent n-gram analysis for training-free detection of gpt-generated text. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenR… find
  64. Rowan Zellers, Ari Holtzman, Hannah Rashkin, Yonatan Bisk, Ali Farhadi, Franziska Roesner, and Yejin Choi. 2019. Defending against neural fake news. In Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, December 8-14,…
  65. Xiang Zhang, Junbo Jake Zhao, and Yann LeCun. 2015. Character-level convolutional networks for text classification. In Advances in Neural Information Processing Systems 28: Annual Conference on Neural Information Processing Systems 2015, December 7-12, 2015, Montreal, Quebec, Canada, pages 649–657.
  66. Yanzhao Zhang, Mingxin Li, Dingkun Long, Xin Zhang, Huan Lin, Baosong Yang, Pengjun Xie, An Yang, Dayiheng Liu, Junyang Lin, Fei Huang, and Jingren Zhou. 2025. Qwen3 embedding: Advancing text embedding and reranking through foundation models. arXiv preprint arXiv:2506.05176.
  67. Hongyi Zhou, Jin Zhu, Pingfan Su, Kai Ye, Ying Yang, Shakeel A O. B. Gavioli-Akilagun, and Chengchun Shi. 2025. Adadetectgpt: Adaptive detection of llm-generated text with statistical guarantees. CoRR, abs/2510.01268. — [arXiv:2510.01268](https://arxiv.org/abs/2510.01268) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/xiong-2026-verifiable-llm-generated-text-detection/