Elamur Research⌕
☰
home / papers / 2026 / Reasoning-Aware AIGC Detection via Alignment and Reinforcement

Reasoning-Aware AIGC Detection via Alignment and Reinforcement

Paper · 2026 · Zhao Wang, Max Xiong, Jianxun Lian et al.

Summary

The paper tackles AI-generated content (AIGC) detection. Older detectors depend on statistical cues or opaque neural classifiers, and they often fail on newer LLMs and on human text that AI has only polished. The authors make two contributions. The first is AIGC-text-bank, a parallel corpus in which each pre-ChatGPT human document is paired with two kinds of counterpart. AI-Native texts are generated from a GPT-4o summary of the original's topic and style. AI-Polish texts are produced by having a model refine the human original. The second contribution is REVEAL, a detector that writes out its reasoning before it gives a label.

Training happens in two phases. Supervised fine-tuning uses rationales from OpenAI o3, which is shown each text with its true label and asked to explain the label after the fact. Reinforcement learning then improves consistency, cuts hallucinated reasoning and lets the model surpass its teacher. On five benchmarks, the authors report that REVEAL outperforms general-purpose LLMs as well as opaque, unexplained detectors, for two-way and three-way labeling alike, and that it stays robust when domains shift or inputs are adversarial. This matters because the detector's judgments come with evidence that a human can check.

Key points

  • AIGC-text-bank composition:
    • 66,979 Human documents (about 22.5M tokens)
    • 699,052 AI-Native samples (about 195.4M tokens)
    • 732,248 AI-Polish samples (about 205.6M tokens)
    • 10 domains, including Wiki, arXiv, Reddit, news, essays and speeches
    • 12 generator LLMs, including GPT-5, Grok-4, DeepSeek R1, Llama 3.3, Qwen 3, Phi-4 and GPT-2
  • Data controls:
    • Human texts all predate November 30, 2022, so they cannot contain ChatGPT output.
    • AI outputs are matched in length to their human references.
    • For 20% of samples, the generator was prompted to imitate human writing, which makes those samples harder to detect.
  • Supervised fine-tuning:
    • Outputs follow a <think>…</think><answer>…</answer> format.
    • An outcome-weighted loss keeps the long reasoning traces from drowning out the final label.
  • Reinforcement learning:
    • Training uses the DAPO algorithm.
    • Only samples whose rewards vary across sampled outputs are kept, so training concentrates on uncertain cases.
  • REVEAL-Fast variant:
    • It skips the reasoning step to produce a calibrated "AIGC score," where 0 means Human, 0.5 AI-Polish and 1 AI-Native.
    • The score correlates strongly with accuracy, whereas the full model gives overconfident probabilities.
    • It also supports judging authorship paragraph by paragraph.

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

🏢Microsoft Research AsiaBeijing · China73 works · 60,539 citations🎓Renmin University of ChinaBeijing · China114 works · 16,118 citations🎓Duke UniversityDurham · United States28 works · 15,077 citations

 Fields, topics and concepts

📚 Shelves 5The 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 worksComputer Vision and Pattern Recognition · 2,869 works

🧭 Topics 1Research subjects and languages the work is about.Reasoning · 1,426 works

Cite

Cite as Zhao Wang, Max Xiong, Jianxun Lian and Zhicheng Dou. 2026. Reasoning-Aware AIGC Detection via Alignment and Reinforcement. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.1043

@inproceedings{wang2026reasoning,
  title = {{Reasoning-Aware AIGC Detection via Alignment and Reinforcement}},
  author = {Zhao Wang, Max Xiong, Jianxun Lian and Zhicheng Dou},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1043},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1043}
}

References

34 references, 17 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Marah Abdin, Jyoti Aneja, Harkirat Behl, Sébastien Bubeck, Ronen Eldan, Suriya Gunasekar, Michael Harrison, Russell J Hewett, Mojan Javaheripi, Piero Kauffmann, and 1 others. 2024. Phi-4 technical report. arXiv preprint arXiv:2412.08905.
  2. Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and 1 others. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774.
  3. Guangsheng Bao, Yanbin Zhao, Zhiyang Teng, Linyi Yang, and Yue Zhang. 2023. Fast-detectgpt: Efficient zero-shot detection of machine-generated text via conditional probability curvature. In The Twelfth International Conference on Learning Representations.
  4. Janek Bevendorff, Daryna Dementieva, Maik Fröbe, Bela Gipp, André Greiner-Petter, Jussi Karlgren, Maximilian Mayerl, Preslav Nakov, Alexander Panchenko, Martin Potthast, and 1 others. 2025. Overview of pan 2025: Generative ai detection, multilingual text detoxification, multi-author writing style an… find
  5. Jiaqi Chen, Xiaoye Zhu, Tianyang Liu, Ying Chen, Chen Xinhui, Yiwen Yuan, Chak Tou Leong, Zuchao Li, Long Tang, Lei Zhang, and 1 others. 2025. Imitate before detect: Aligning machine stylistic preference for machine-revised text detection. In Proceedings of the AAAI Conference on Artificial Intellig… find
  6. Vincenzo Ciancaglini, Craig Gibson, David Sancho, Odhran McCarthy, Maria Eira, Philipp Amann, and Aglika Klayn. 2020. Malicious uses and abuses of artificial intelligence. Trend Micro Research, pages 4–79. find
  7. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, and 1 others. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.
  8. Biyang Guo, Xin Zhang, Ziyuan Wang, Minqi Jiang, Jinran Nie, Yuxuan Ding, Jianwei Yue, and Yupeng Wu. 2023. How close is chatgpt to human experts? comparison corpus, evaluation, and detection. arXiv preprint arXiv:2301.07597.
  9. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Peiyi Wang, Qihao Zhu, Runxin Xu, Ruoyu Zhang, Shirong Ma, Xiao Bi, and 1 others. 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948.
  10. 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 Proceedings of the 41st International Conference on Machine Learning, ICML’2… find
  11. Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, and 1 others. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276.
  12. 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.
  13. 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. Advances in neural information processing systems, 36:27469–27500.
  14. Thomas Lavergne, Tanguy Urvoy, and François Yvon. 2008. Detecting fake content with relative entropy scoring. Pan, 8(27-31):4. find
  15. 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. Preprint, arXiv:1907.11692.
  16. 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, pages 24950–24962. PMLR.
  17. OpenAI. 2024a. GPT-3.5 Turbo fine-tuning and API updates. find
  18. OpenAI. 2024b. GPT-4o mini: advancing cost-efficient intelligence. find
  19. OpenAI. 2024c. Introducing OpenAI o3 and o4-mini. find
  20. Mike Perkins. 2023. Academic integrity considerations of ai large language models in the post-pandemic era: Chatgpt and beyond. Journal of University Teaching and Learning Practice, 20(2):1–24. find
  21. Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, and 25 others. 2025. Qwen2.5 technical report. Preprint, arXiv:2412.15115. find
  22. Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, and 1 others. 2019. Language models are unsupervised multitask learners. OpenAI blog, 1(8):9. find
  23. Aaditya Singh, Adam Fry, Adam Perelman, Adam Tart, Adi Ganesh, Ahmed El-Kishky, Aidan McLaughlin, Aiden Low, AJ Ostrow, Akhila Ananthram, and 1 others. 2025. Openai gpt-5 system card. arXiv preprint arXiv:2601.03267.
  24. Irene Solaiman, Miles Brundage, Jack Clark, Amanda Askell, Ariel Herbert-Voss, Jeff Wu, Alec Radford, Gretchen Krueger, Jong Wook Kim, Sarah Kreps, and 1 others. 2019. Release strategies and the social impacts of language models. arXiv preprint arXiv:1908.09203.
  25. Zhenpeng Su, Xing Wu, Wei Zhou, Guangyuan Ma, and Songlin Hu. 2023. Hc3 plus: A semantic-invariant human chatgpt comparison corpus. arXiv preprint arXiv:2309.02731. — [arXiv:2309.02731](https://arxiv.org/abs/2309.02731) find
  26. Qwen Team. 2025. Qwq-32b: Embracing the power of reinforcement learning. find
  27. Adaku Uchendu, Zeyu Ma, Thai Le, Rui Zhang, and Dongwon Lee. 2021. Turingbench: A benchmark environment for turing test in the age of neural text generation. In Findings of the association for computational linguistics: EMNLP 2021, pages 2001–2016.
  28. Leandro von Werra, Younes Belkada, Lewis Tunstall, Edward Beeching, Tristan Thrush, Nathan Lambert, Shengyi Huang, Kashif Rasul, and Quentin Gallouédec. 2020. Trl: Transformer reinforcement learning. https://github.com/huggingface/trl. ↗
  29. Yuxia Wang, Jonibek Mansurov, Petar Ivanov, Jinyan Su, Artem Shelmanov, Akim Tsvigun, Chenxi Whitehouse, Osama Mohammed Afzal, Tarek Mahmoud, Toru Sasaki, and 1 others. 2024. M4: Multi-generator, multi-domain, and multi-lingual black-box machine-generated text detection. In Proceedings of the 18th C… find
  30. Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pier-ric Cistac, Tim Rault, Rémi Louf, Morgan Funtowicz, and 1 others. 2020. Transformers: State-of-the-art natural language processing. In Proceedings of the 2020 conference on empirical methods in natural lan…
  31. Junchao Wu, Runzhe Zhan, Derek F Wong, Shu Yang, Xinyi Yang, Yulin Yuan, and Lidia S Chao. 2024. Detectrl: Benchmarking llm-generated text detection in real-world scenarios. Advances in Neural Information Processing Systems, 37:100369–100401. xAI. 2024. Grok 4.1. find
  32. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and 1 others. 2025. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  33. Junyan Ye, Baichuan Zhou, Zilong Huang, Junan Zhang, Tianyi Bai, Hengrui Kang, Jun He, Honglin Lin, Zihao Wang, Tong Wu, and 1 others. 2025. Loki: A comprehensive synthetic data detection benchmark using large multimodal models. ICLR. find
  34. Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Weinan Dai, Tiantian Fan, Gaohong Liu, Lingjun Liu, and 1 others. 2025. Dapo: An open-source llm reinforcement learning system at scale. arXiv preprint arXiv:2503.14476.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-reasoning-aware-aigc-detection-alignment/