Elamur Research⌕
☰
home / papers / 2026 / Explainable Disentangled Representation Learning for Generalizable Authorship Attribution in the Era of Generative AI

Explainable Disentangled Representation Learning for Generalizable Authorship Attribution in the Era of Generative AI

Paper · 2026 · Hieu Man, Van-Cuong Pham, Nghia Trung Ngo et al.

Summary

Authorship attribution models often confuse an author's writing style with the topics that author tends to cover. This topic confusion hurts performance when the domain changes. The authors propose EAVAE (Explainable Authorship Variational Autoencoder), which keeps style and content apart by giving each its own encoder.

Training has two stages:

  • Contrastive pretraining. An LLM-based style encoder, with bidirectional attention enabled, is pretrained using supervised contrastive learning. Hard negatives are mined with BM25.
  • VAE finetuning. Separate style and content encoders produce Gaussian latents, and a shared reconstructor rebuilds the document from them. A discriminator is trained adversarially to judge whether two style vectors come from the same author, or two content vectors from the same source. It also writes a natural-language justification for each judgment. Reconstruction and discrimination share one generator that uses hybrid prompting.

EAVAE reports state-of-the-art attribution results on Amazon Reviews, PAN21 and HRS. It also performs strongly on AI-generated text detection on M4 with very little labeled data. This matters because it gives style representations that transfer across domains and come with explanations, which is relevant to forensic work and to detecting LLM-written text.

Key points

  • Evaluation: Authorship attribution is framed as ranked retrieval by cosine similarity and evaluated with MRR and R@8. Benchmarks are Amazon Reviews, PAN21 and the HRS corpus from the IARPA HIATUS program, plus M4 for machine-text detection.
  • Ablations on the discriminator:
    • The hybrid-prompting discriminator outperforms an MLP discriminator (MRR 45.5, R@8 65.4).
    • It also outperforms a soft-prompt-only variant (MRR 43.3, R@8 66.1), beating it by 4.0 MRR points.
    • Removing the component that enforces style-content independence costs 1.9 MRR and 6.2 R@8.
  • Objectives: The VAE loss uses separate KL weights for style and content (β_s, β_c) and assumes the two latents are independent given the document.
  • Resources: Code is released at github.com/hieum98/avae, and document-level authorship datasets are published as a Hugging Face collection (Hieuman).
  • Limitations: The authors note that explanation quality depends on the underlying LLM, and that the method has not yet been extended to multi-author texts. Multilingual extension is proposed as future work, which is relevant for possible Ukrainian applications.

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

🎓University of OregonEugene · United States23 works · 3,272 citations🏢Adobe SystemsSan Jose · United States35 works · 3,711 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksPsychology · 1,552 worksSocial Psychology · 306 works

Cite

Cite as Hieu Man, Van-Cuong Pham, Nghia Trung Ngo, Franck Dernoncourt and Thien Huu Nguyen. 2026. Explainable Disentangled Representation Learning for Generalizable Authorship Attribution in the Era of Generative AI. 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.2018

@inproceedings{man2026explainable,
  title = {{Explainable Disentangled Representation Learning for Generalizable Authorship Attribution in the Era of Generative AI}},
  author = {Hieu Man, Van-Cuong Pham, Nghia Trung Ngo, Franck Dernoncourt and Thien Huu Nguyen},
  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.2018},
  url = {https://doi.org/10.18653/v1/2026.acl-long.2018}
}

References

41 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Bo Ai, Yuchen Wang, Yugin Tan, and Samson Tan. 2022. Whodunit? learning to contrast for authorship attribution. In Proceedings of the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Conference on Natural Language Processing… find
  2. Malik Altakrori, Jackie Chi Kit Cheung, and Benjamin CM Fung. 2021. The topic confusion task: A novel evaluation scenario for authorship attribution. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 4242–4256. find
  3. Parishad BehnamGhader, Vaibhav Adlakha, Marius Mosbach, Dzmitry Bahdanau, Nicolas Chapados, and Siva Reddy. 2024. Llm2vec: Large language models are secretly powerful text encoders. Preprint, arXiv:2404.05961.
  4. Janek Bevendorff, Bilal Ghanem, Anastasia Giachanou, Mike Kestemont, Enrique Manjavacas, Ilia Markov, Maximilian Mayerl, Martin Potthast, Francisco Rangel, Paolo Rosso, Günther Specht, Efstathios Stamatatos, Benno Stein, Matti Wiegmann, and Eva Zangerle. 2020. Overview of pan 2020: Authorship verifi… find
  5. Benedikt Boenninghoff, Steffen Hessler, Dorothea Kolossa, and Robert M Nickel. 2019. Explainable authorship verification in social media via attention-based similarity learning. In 2019 IEEE International Conference on Big Data (Big Data), pages 36–45. IEEE. find
  6. Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, Pascal Germain, Hugo Larochelle, François Laviolette, Mario Marchand, and Victor Lempitsky. 2016. Domain-adversarial training of neural networks. Preprint, arXiv:1505.07818. — [arXiv:1505.07818](https://arxiv.org/abs/1505.07818) find
  7. Pengzhi Gao, Liwen Zhang, Zhongjun He, Hua Wu, and Haifeng Wang. 2023. Learning multilingual sentence representations with cross-lingual consistency regularization. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 243–262, Singapore. As… find
  8. Hieu Man, Nghia Trung Ngo, Viet Dac Lai, Ryan A. Rossi, Franck Dernoncourt, and Thien Huu Nguyen. 2025. LUSIFER: Language universal space integration for enhanced multilingual text embedding models. In Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Informat… find
  9. Julian Hazell. 2023. Spear phishing with large language models. Preprint, arXiv:2305.06972. — [arXiv:2305.06972](https://arxiv.org/abs/2305.06972) find
  10. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. Lora: Low-rank adaptation of large language models. Preprint, arXiv:2106.09685.
  11. Hieu Man and Huu Thien Nguyen. 2024. Counterfactual augmentation for robust authorship representation learning. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR ’24, page 2347–2351, New York, NY, USA. Association for Computing… find
  12. IARPA. HIATUS — iarpa.gov. https://www.iarpa.gov/research-programs/hiatus. [Accessed 23-09-2025]. ↗
  13. Niklas Muennighoff, Hongjin Su, Liang Wang, Nan Yang, Furu Wei, Tao Yu, Amanpreet Singh, and Douwe Kiela. 2025. Generative representational instruction tuning. Preprint, arXiv:2402.09906.
  14. Hemanth Kandula, Chak Fai Li, Haoling Qiu, Damianos Karakos, Hieu Man, Thien Huu Nguyen, and Brian Ulicny. 2025. BBN-U.Oregon’s ALERT system at GenAI content detection task 3: Robust authorship style representations for cross-domain machine-generated text detection. In Proceedings of the 1st Worksho… find
  15. Jianmo Ni, Jiacheng Li, and Julian McAuley. 2019. Justifying recommendations using distantly-labeled reviews and fine-grained aspects. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (…
  16. Prannay Khosla, Piotr Teterwak, Chen Wang, Aaron Sarna, Yonglong Tian, Phillip Isola, Aaron Maschinot, Ce Liu, and Dilip Krishnan. 2021. Supervised contrastive learning. Preprint, arXiv:2004.11362. — [arXiv:2004.11362](https://arxiv.org/abs/2004.11362) find
  17. Geon Yeong Park and Sang Wan Lee. 2021. Information-theoretic regularization for multi-source domain adaptation. In Proceedings of the IEEE/CVF international conference on computer vision, pages 9214–9223. find
  18. Diederik P Kingma and Max Welling. 2022. Auto-encoding variational bayes. Preprint, arXiv:1312.6114.
  19. Qwen Qwen Team. 2025. Qwq-32b: Embracing the power of reinforcement learning. find
  20. Varada Kolhatkar, Hanhan Wu, Luca Cavasso, Emilie Francis, Kavan Shukla, and Maite Taboada. 2020. The sfu opinion and comments corpus: A corpus for the analysis of online news comments. Corpus pragmatics, 4(2):155–190. find
  21. Zehan Li, Xin Zhang, Yanzhao Zhang, Dingkun Long, Pengjun Xie, and Meishan Zhang. 2023. Towards general text embeddings with multi-stage contrastive learning. arXiv preprint arXiv:2308.03281.
  22. Ilya Loshchilov and Frank Hutter. 2019. Decoupled weight decay regularization. Preprint, arXiv:1711.05101. find
  23. Hieu Man, Nghia Trung Ngo, Franck Dernoncourt, and Thien Huu Nguyen. 2024. ULLME: A unified framework for large language model embeddings with generation-augmented learning. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 230–23… find
  24. Abhinav Ramesh Kashyap, Devamanyu Hazarika, Min-Yen Kan, Roger Zimmermann, and Soujanya Poria. 2022. So different yet so alike! constrained unsupervised text style transfer. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 416–… find
  25. Rafael A. Rivera-Soto, Olivia Elizabeth Miano, Juanita Ordonez, Barry Y. Chen, Aleem Khan, Marcus Bishop, and Nicholas Andrews. 2021. Learning universal authorship representations. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 913–919, Online and Pu… find
  26. Jitkapat Sawatphol, Nonthakit Chaiwong, Can Udomcharoenchaikit, and Sarana Nutanong. 2022. Topic-regularized authorship representation learning. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 1076–1082, Abu Dhabi, United Arab Emirates. Association fo… find
  27. Yanir Seroussi, Ingrid Zukerman, and Fabian Bohnert. 2014. Authorship attribution with topic models. Computational Linguistics, 40(2):269–310.
  28. Rafael Rivera Soto, Kailin Koch, Aleem Khan, Barry Chen, Marcus Bishop, and Nicholas Andrews. 2024. Few-shot detection of machine-generated text using style representations. Preprint, arXiv:2401.06712. — [arXiv:2401.06712](https://arxiv.org/abs/2401.06712) find
  29. Efstathios Stamatatos. 2009. A survey of modern authorship attribution methods. J. Am. Soc. Inf. Sci. Technol., 60(3):538–556. find
  30. Efstathios Stamatatos. 2017. Authorship attribution using text distortion. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, Long Papers, pages 1138–1149, Valencia, Spain. Association for Computational Linguistics. find
  31. Ariel Stolerman, Rebekah Overdorf, Sadia Afroz, and Rachel Greenstadt. 2014. Breaking the closed-world assumption in stylometric authorship attribution. In IFIP International Conference on Digital Forensics, pages 185–205. Springer. find
  32. Justin Anthony Stover, Yaron Winter, Moshe Koppel, and Mike Kestemont. 2016. Computational authorship verification method attributes a new work to a major 2nd century a frican author. Journal of the Association for Information Science and Technology, 67(1):239–242. find
  33. Mengting Wan, Rishabh Misra, Ndapa Nakashole, and Julian McAuley. 2019. Fine-grained spoiler detection from large-scale review corpora. arXiv preprint arXiv:1905.13416. find
  34. John Wieting, Jonathan Clark, William Cohen, Graham Neubig, and Taylor Berg-Kirkpatrick. 2023. Beyond contrastive learning: A variational generative model for multilingual retrieval. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), p… find
  35. An Yang, Baosong Yang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Zhou, Chengpeng Li, Chengyuan Li, Dayiheng Liu, Fei Huang, Guanting Dong, Haoran Wei, Huan Lin, Jialong Tang, Jialin Wang, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Ma, and 43 others. 2024. Qwen2 technical report. Preprint, arXiv:240… find
  36. 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.
  37. Jiarui Zhang and Yonghua Zhu. 2021. Meta-path guided heterogeneous graph neural network for dish recommendation system. In Journal of Physics: Conference Series, volume 1883, page 012102. IOP Publishing. find
  38. Yukun Zhu, Ryan Kiros, Rich Zemel, Ruslan Salakhutdinov, Raquel Urtasun, Antonio Torralba, and Sanja Fidler. 2015. Aligning books and movies: Towards story-like visual explanations by watching movies and reading books. In The IEEE International Conference on Computer Vision (ICCV).
  39. Yuxia Wang, Jonibek Mansurov, Petar Ivanov, Jinyan Su, Artem Shelmanov, Akim Tsvigun, Chenxi Whitehouse, Osama Mohammed Afzal, Tarek Mahmoud, Toru Sasaki, Thomas Arnold, Alham Fikri Aji, Nizar Habash, Iryna Gurevych, and Preslav Nakov. 2024. M4: Multi-generator, multi-domain, and multilingual black-… find
  40. Anna Wegmann, Marijn Schraagen, and Dong Nguyen. 2022. Same author or just same topic? towards content-independent style representations. In Proceedings of the 7th Workshop on Representation Learning for NLP, pages 249–268, Dublin, Ireland. Association for Computational Linguistics. find
  41. Laura Weidinger, Jonathan Uesato, Maribeth Rauh, Conor Griffin, Po-Sen Huang, John Mellor, Amelia Glaese, Myra Cheng, Borja Balle, Atoosa Kasirzadeh, Courtney Biles, Sasha Brown, Zac Kenton, Will Hawkins, Tom Stepleton, Abeba Birhane, Lisa Anne Hendricks, Laura Rimell, William Isaac, and 4 others. 2… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/man-2026-explainable-disentangled-representation-learning-generalizable/