Elamur Research⌕
☰
home / papers / 2026 / Detecting Hallucinations in Authentic LLM–Human Interactions

Detecting Hallucinations in Authentic LLM–Human Interactions

Summary

Most hallucination detection benchmarks use artificial data. Either an LLM is told outright to produce wrong answers, or queries are taken from existing datasets to imitate real usage. Neither setup shows how hallucinations actually appear when people talk to chatbots. The authors fill this gap with AuthenHallu, a benchmark built only from genuine conversations in LMSYS-Chat-1M.

They first filtered the logs to keep English, two-turn, safe, non-redacted, non-duplicated dialogues of suitable length. They then clustered the remaining conversations and sampled from each cluster in proportion to its size. Finally, they hand-labeled every query–response pair for whether it hallucinates and, if so, which type: conflict with the user input, with earlier context, or with world facts (taxonomy of Zhang et al., 2025).

Roughly a third of the responses contain hallucinations, mostly factual errors, and mathematical queries fail most often. Off-the-shelf LLMs used as detectors and categorizers show some promise but are not yet reliable for real-world use. The benchmark gives a more ecologically valid way to test detection methods.

Key points

  • Size and structure: AuthenHallu has 400 authentic dialogues, each with two turns, for 800 annotated query–response pairs. The source is LMSYS-Chat-1M (conversations logged April–August 2023).
  • Labels: each pair is marked Hallucination or No hallucination. Hallucinated pairs are further typed as Input-conflicting, Context-conflicting or Fact-conflicting.
  • Annotation: three annotators labeled the same 200 dialogues, reaching Fleiss's Kappa = 0.591. One annotator then labeled the other 200.
  • Statistics:
    • 31.4% of pairs contain hallucinations.
    • Fact-conflicting cases make up 62.5% of all hallucinations.
    • The "Math & Number Problems" topic has the highest hallucination rate, 60.0%.
  • Experiments and availability: vanilla LLMs were prompted for both detection and categorization, with single-pair and in-context (full dialogue) prompts. Results fall short of practical needs. Data and code are released at github.com/TAI-HAMBURG/AuthenHallu.

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

Fields: Linguistics & languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/5pykihiz52tk

@inproceedings{ren2026detecting,
  title = {{Detecting Hallucinations in Authentic LLM–Human Interactions}},
  author = {Ren, Yujie and Gruhlke, Niklas and Lauscher, Anne},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/5pykihiz52tk},
  url = {https://doi.org/10.63317/5pykihiz52tk}
}

References

38 references, 10 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Orlando Ayala and Patrice Bechard. 2024. Reducing hallucination in structured outputs via retrieval-augmented generation. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 6: Industry Track), page… find
  2. Forrest Sheng Bao, Miaoran Li, Renyi Qu, Ge Luo, Erana Wan, Yujia Tang, Weisi Fan, Manveer Singh Tamber, Suleman Kazi, Vivek Sourabh, Mike Qi, Ruixuan Tu, Chenyu Xu, Matthew Gonzales, Ofer Mendelevitch, and Amin Ahmad. 2025. FaithBench: A diverse hallucination benchmark for summarization by Modern L… find
  3. Kedi Chen, Qin Chen, Jie Zhou, He Yishen, and Liang He. 2024a. DiaHalu: A dialogue-level hallucination evaluation benchmark for large language models. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 9057–9079, Miami, Florida, USA. Association for Computational Linguis… find
  4. Shiqi Chen, Yiran Zhao, Jinghan Zhang, I-Chun Chern, Siyang Gao, Pengfei Liu, and Junxian He. 2023. FELM: Benchmarking factuality evaluation of large language models. In Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track. find
  5. Xiang Chen, Duanzheng Song, Honghao Gui, Chenxi Wang, Ningyu Zhang, Yong Jiang, Fei Huang, Chengfei Lyu, Dan Zhang, and Huajun Chen. 2024b. Factchd: benchmarking fact-conflicting hallucination detection. In Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, IJ… find
  6. Zheng Chu, Jingchang Chen, Qianglong Chen, Weijiang Yu, Haotian Wang, Ming Liu, and Bing Qin. 2024. TimeBench: A comprehensive evaluation of temporal reasoning abilities in large language models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Lo… find
  7. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021. Training verifiers to solve math word problems. find our card
  8. Joseph L Fleiss. 1971. Measuring nominal scale agreement among many raters. Psychological bulletin, 76(5):378. find
  9. Aaron Grattafiori, Abhimanyu Dubey, ..., and Zhiyu Ma. 2024. The llama 3 herd of models. find
  10. Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. 2021. Measuring mathematical problem solving with the math dataset. find our card
  11. Lei Huang, Weijiang Yu, Weitao Ma, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qianglong Chen, Weihua Peng, Xiaocheng Feng, Bing Qin, and Ting Liu. 2025. A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions. ACM Trans. Inf. Syst., 43(2). find our card
  12. Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix,… find
  13. Adam Tauman Kalai, Ofir Nachum, Santosh S. Vempala, and Edwin Zhang. 2025. Why language models hallucinate. find
  14. J. Richard Landis and Gary G. Koch. 1977. The measurement of observer agreement for categorical data. Biometrics, 33(1):159174. find
  15. Junyi Li, Xiaoxue Cheng, Xin Zhao, Jian-Yun Nie, and Ji-Rong Wen. 2023. HaluEval: A large-scale hallucination evaluation benchmark for large language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 6449–6464, Singapore. Association for Computa… find our card
  16. Bill Yuchen Lin, Yuntian Deng, Khyathi Raghavi Chandu, Faeze Brahman, Abhilasha Ravichander, Valentina Pyatkin, Nouha Dziri, Ronan Le Bras, and Yejin Choi. 2024. Wildbench: Benchmarking llms with challenging tasks from real users in the wild. CoRR, abs/2406.04770. find
  17. Wen Luo, Tianshu Shen, Wei Li, Guangyue Peng, Richeng Xuan, Houfeng Wang, and Xi Yang. 2024. Halludial: A large-scale benchmark for automatic dialogue-level hallucination evaluation. find
  18. Potsawee Manakul, Adian Liusie, and Mark JF Gales. 2023. Selfcheckgpt: Zero-resource black-box hallucination detection for generative large language models. arXiv preprint arXiv:2303.08896. find
  19. Ariana Martino, Michael Iannelli, and Coleen Truong. 2023. Knowledge injection to counter large language model (llm) hallucination. In The Semantic Web: ESWC 2023 Satellite Events: Hersonissos, Crete, Greece, May 28 - June 1, 2023, Proceedings, page 182–185, Berlin, Heidelberg. Springer-Verlag. find
  20. Abhika Mishra, Akari Asai, Vidhisha Balachandran, Yizhong Wang, Graham Neubig, Yulia Tsvetkov, and Hannaneh Hajishirzi. 2024. Fine-grained hallucination detection and editing for language models. In First Conference on Language Modeling. find
  21. OpenAI, Aaron Hurst, ..., and Yury Malkov. 2024. Gpt-4o system card. find our card
  22. Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks. find
  23. Peter J Rousseeuw. 1987. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. Journal of computational and applied mathematics, 20:53–65. find
  24. Gerard Salton, Anita Wong, and Chung-Shu Yang. 1975. A vector space model for automatic indexing. Communications of the ACM, 18(11):613– 620. find
  25. Gemma Team, Aishwarya Kamath, ..., and Léonard Hussenot. 2025. Gemma 3 technical report. find
  26. Qwen Team. 2024. Qwen2.5: A party of foundation models. find
  27. Saad Obaid ul Islam, Anne Lauscher, and Goran Glavaš. 2025. How much do llms hallucinate across languages? on multilingual estimation of llm hallucination in the wild. find
  28. Yuqing Wang and Yun Zhao. 2024. TRAM: Benchmarking temporal reasoning for large language models. In Findings of the Association for Computational Linguistics: ACL 2024, pages 6389– 6415, Bangkok, Thailand. Association for Computational Linguistics. find
  29. Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Rémi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mariama Drame, Quentin Lhoe… find our card
  30. An Yang, Anfeng Li, ..., and Zihan Qiu. 2025. Qwen3 technical report. find our card
  31. Shiping Yang, Renliang Sun, and Xiaojun Wan. 2023. A new benchmark and reverse validation method for passage-level hallucination detection. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 3898–3908, Singapore. Association for Computational Linguistics. find
  32. Yue Zhang, Yafu Li, Leyang Cui, Deng Cai, Lemao Liu, Tingchen Fu, Xinting Huang, Enbo Zhao, Yu Zhang, Yulong Chen, et al. 2025. Siren’s song in the ai ocean: A survey on hallucination in large language models. Computational Linguistics, pages 1–46. find our card
  33. Wenting Zhao, Tanya Goyal, Yu Ying Chiu, Liwei Jiang, Benjamin Newman, Abhilasha Ravichander, Khyathi Chandu, Ronan Le Bras, Claire Cardie, Yuntian Deng, and Yejin Choi. 2024a. Wildhallucinations: Evaluating long-form factuality in llms with real-world entity queries. find
  34. Wenting Zhao, Xiang Ren, Jack Hessel, Claire Cardie, Yejin Choi, and Yuntian Deng. 2024b. Wildchat: 1m chatGPT interaction logs in the wild. In The Twelfth International Conference on Learning Representations. find our card
  35. Yukun Zhao, Lingyong Yan, Weiwei Sun, Guoliang Xing, Chong Meng, Shuaiqiang Wang, Zhicong Cheng, Zhaochun Ren, and Dawei Yin. 2024c. Knowing what LLMs DO NOT know: A simple yet effective self-detection method. In Proceedings of the 2024 Conference of the North American Chapter of the Association for… find
  36. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Tianle Li, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zhuohan Li, Zi Lin, Eric. P Xing, Joseph E. Gonzalez, Ion Stoica, and Hao Zhang. 2023. Lmsys-chat-1m: A large-scale real-world llm conversation dataset. find our card
  37. Zhiying Zhu, Yiming Yang, and Zhiqing Sun. 2025. Halueval-wild: Evaluating hallucinations of language models in the wild. In ICLR 2025 Workshop on Building Trust in Language Models and Applications. 0.30 225 200 175 150 125 100 75 50 find
  38. Silhouette Score 0.29 Inertia (x1000) 0.28 0.27 0.26 Number of Clusters Number of Clusters (a) Silhouette and inertia for clustering on first-pair queries. 0.295 Silhouette Score Inertia (x1000) 0.290 0.285 0.280 0.275 Number of Clusters Number of Clusters (b) Silhouette and inertia for clustering o… find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/ren-2026-detecting-hallucinations-authentic-llm-human/