Elamur Research⌕
☰
home / papers / 2026 / Thinking Alignment of Scenario-Oriented User Simulation

Thinking Alignment of Scenario-Oriented User Simulation

Paper · 2026 · Xiaoting Wu, Yi Huang, Chunyang Gao et al.

Summary

LLM-based user simulators, whether they prompt a model to role-play or fine-tune it on real user turns, usually reproduce only what users type. They ignore the reasoning that produces those messages. This paper tries to make simulated users think more like real people. The authors model the user side of a dialogue in three layers: a scenario setting that holds for the whole conversation (a background description plus the user's motivation), a hidden thought at each turn, and the visible utterance. They start from real human–LLM chat logs and use a strong LLM to infer scenarios and reconstruct the hidden reasoning, guided by a four-dimensional cognitive framework. The resulting dataset, LMSYS-UserThinking, covers roughly 51k conversations. They train simulators called ThinkingUS on several base models and test them offline and in live dialogues. ThinkingUS outperforms both prompt-based role-play and SFT on utterances alone. More faithful simulators can supply better synthetic multi-turn data, RL environments and interactive evaluation for assistant models.

Key points

  • LMSYS-UserThinking: about 51k human–LLM dialogues, augmented with a scenario description, user motivation and per-turn thoughts. The paper draws on LMSYS-Chat-1M; WildChat is also listed among its data sources.
  • Two kinds of reconstructed thought:
    • In-process thoughts connect the assistant's last reply to the user's next message.
    • Terminal thoughts decide when the user's goal is met and the chat should end.
    • Both are framed by the cognitive framework of Kargupta et al. (2025b), which covers metacognitive control and reasoning operations such as verification and backtracking.
  • Online evaluation: thinking-enabled simulators score roughly 70–80% higher on goal disclosure than non-thinking ones. They also introduce more useful details that were not in the scenario.
  • Case study: prompted "thinking" simulators tend to slip into a problem-solver role and reveal all requirements early. ThinkingUS is more concise, moves from easy to hard questions, and stops once it is satisfied.
  • Limitations: the thoughts are inferred from dialogue context and may differ from users' actual cognition. Source data may also bias the coverage of scenarios.

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

🎓Tsinghua UniversityBeijing · China333 works · 87,019 citations

 Fields, topics and concepts

📚 Shelves 4The 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 works

Cite

Cite as Xiaoting Wu, Yi Huang, Chunyang Gao, Mengfei Guo, Jingyu Yao and Junlan Feng. 2026. Thinking Alignment of Scenario-Oriented User Simulation. 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.1989

@inproceedings{wu2026thinking,
  title = {{Thinking Alignment of Scenario-Oriented User Simulation}},
  author = {Xiaoting Wu, Yi Huang, Chunyang Gao, Mengfei Guo, Jingyu Yao and Junlan Feng},
  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.1989},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1989}
}

References

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

  1. Kaustubh Dhole. 2024. KAUCUS - knowledgeable user simulators for training large language models. In Proceedings of the 1st Workshop on Simulating Conversational Intelligence in Chat (SCI-CHAT 2024), pages 53–65, St. Julians, Malta. Association for Computational Linguistics. find
  2. Ning Ding, Yulin Chen, Bokai Xu, Yujia Qin, Shengding Hu, Zhiyuan Liu, Maosong Sun, and Bowen Zhou. 2023. Enhancing chat language models by scaling high-quality instructional conversations. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapo…
  3. Vardhan Dongre, Xiaocheng Yang, Emre Can Acikgoz, Suvodip Dey, Gokhan Tur, and Dilek Hakkani-Tur. 2025. ReSpAct: Harmonizing reasoning, speaking, and acting towards building large language model-based conversational AI agents. In Proceedings of the 15th International Workshop on Spoken Dialogue Syst… find
  4. Rafael Ferreira, David Semedo, and Joao Magalhaes. 2024. Multi-trait user simulation with adaptive decoding for conversational task assistants. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 16105–16130, Miami, Florida, USA. Association for Computational Linguistics. find
  5. Christine Herlihy, Jennifer Neville, Tobias Schnabel, and Adith Swaminathan. 2024. On overcoming miscalibrated conversational priors in LLM-based chatbots. In The 40th Conference on Uncertainty in Artificial Intelligence. find
  6. Friederike Holderried, Christian Stegemann-Philipps, Anne Herrmann-Werner, Teresa Festl-Wietek, Martin Holderried, Carsten Eickhoff, Moritz Mahling, and 1 others. 2024. A language model–powered simulated patient with automated feedback for history taking: Prospective study. JMIR Medical Education, 1… find
  7. EunJeong Hwang, Bodhisattwa Majumder, and Niket Tandon. 2023. Aligning language models to user opinions. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 5906– 5919, Singapore. Association for Computational Linguistics.
  8. Jonathan Ivey, Shivani Kumar, Jiayu Liu, Hua Shen, Sushrita Rakshit, Rohan Raju, Haotian Zhang, Aparna Ananthasubramaniam, Junghwan Kim, Bowen Yi, Dustin Wright, Abraham Israeli, Anders Giovanni Møller, Lechen Zhang, and David Jurgens. 2024. Real or robotic? assessing whether llms accurately simulat… — [arXiv:2409.08330](https://arxiv.org/abs/2409.08330) find
  9. Jiaming Ji, Boyuan Chen, Hantao Lou, Donghai Hong, Borong Zhang, Xuehai Pan, Tianyi Qiu, Juntao Dai, and Yaodong Yang. 2024. Aligner: Efficient alignment by learning to correct. Advances in Neural Information Processing Systems 37. find
  10. Suhong Moon, Marwa Abdulhai, Minwoo Kang, Joseph Suh, Widyadewi Soedarmadji, Eran Kohen Behar, and David M. Chan. 2024. Virtual personas for language models via an anthology of backstories. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 19864–19897,… find
  11. Priyanka Kargupta, Ishika Agarwal, Tal August, and Jiawei Han. 2025a. Tree-of-debate: Multi-persona debate trees elicit critical thinking for scientific comparative analysis. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 293… find
  12. Priyanka Kargupta, Ishika Agarwal, Dilek Hakkani Tur, and Jiawei Han. 2024. Instruct, not assist: LLM-based multi-turn planning and hierarchical questioning for socratic code debugging. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 9475–9495, Miami, Florida, USA. As… find
  13. Priyanka Kargupta, Shuyue Stella Li, Haocheng Wang, Jinu Lee, Shan Chen, Orevaoghene Ahia, Dean Light, Thomas L. Griffiths, Max Kleiman-Weiner, Jiawei Han, Asli Celikyilmaz, and Yulia Tsvetkov. 2025b. Cognitive foundations for reasoning and their manifestation in llms. Preprint, arXiv:2511.16660. — [arXiv:2511.16660](https://arxiv.org/abs/2511.16660) find
  14. Aobo Kong, Shiwan Zhao, Hao Chen, Qicheng Li, Yong Qin, Ruiqi Sun, Xin Zhou, Enzhi Wang, and Xiaohang Dong. 2024a. Better zero-shot reasoning with role-play prompting. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language… find
  15. Chuyi Kong, Yaxin Fan, Xiang Wan, Feng Jiang, and Benyou Wang. 2024b. PlatoLM: Teaching LLMs in multi-round dialogue via a user simulator. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7841–7863, Bangkok, Thailand. Associati… find
  16. Ruosen Li, Ruochen Li, Barry Wang, and Xinya Du. 2024. IQA-EVAL: automatic evaluation of human-model interactive question answering. In Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver, BC, Canada, December… find
  17. Yajiao Liu, Xin Jiang, Yichun Yin, Yasheng Wang, Fei Mi, Qun Liu, Xiang Wan, and Benyou Wang. 2023. One cannot stand for everyone! leveraging multiple user simulators to train task-oriented dialogue systems. In Proceedings of the 61st Annual Meeting of Tarek Naous, Philippe Laban, Wei Xu, and Jennif… find
  18. Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Bernstein. 2023. Generative agents: Interactive simulacra of human behavior. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology, UIST ’23, New York, NY, USA. Ass… find
  19. Cheng Qian, Zuxin Liu, Akshara Prabhakar, Zhiwei Liu, Jianguo Zhang, Haolin Chen, Heng Ji, Weiran Yao, Shelby Heinecke, Silvio Savarese, Caiming Xiong, and Huan Wang. 2025a. Userbench: An interactive gym environment for user-centric agents. Preprint, arXiv:2507.22034. — [arXiv:2507.22034](https://arxiv.org/abs/2507.22034) find
  20. Cheng Qian, Zuxin Liu, Akshara Prabhakar, Jielin Qiu, Zhiwei Liu, Haolin Chen, Shirley Kokane, Heng Ji, Weiran Yao, Shelby Heinecke, Silvio Savarese, Caiming Xiong, and Huan Wang. 2025b. Userrl: Training interactive user-centric agent via reinforcement learning. Preprint, arXiv:2509.19736. — [arXiv:2509.19736](https://arxiv.org/abs/2509.19736) find
  21. Ivan Sekuli’c, Silvia Terragni, Victor Guimarães, Nghia Khau, Bruna Guedes, Modestas Filipavicius, André Ferreira Manso, and Roland Mathis. 2024. Reliable llm-based user simulator for task-oriented dialogue systems. ArXiv, abs/2402.13374. find
  22. Ryan Shea, Yunan Lu, Liang Qiu, and Zhou Yu. 2025. Sage: A top-down bottom-up knowledge-grounded user simulator for multi-turn agent evaluation. ArXiv, abs/2510.11997.
  23. Dazhen Wan, Zheng Zhang, Qi Zhu, Lizi Liao, and Minlie Huang. 2022. A unified dialogue user simulator for few-shot data augmentation. In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 3788–3799, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics. find
  24. Kuang Wang, Xianfei Li, Shenghao Yang, Li Zhou, Feng Jiang, and Haizhou Li. 2025a. Know you first and be you better: Modeling human-like user simulators via implicit profiles. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 21… find
  25. Lei Wang, Jingsen Zhang, Hao Yang, Zhi-Yuan Chen, Jiakai Tang, Zeyu Zhang, Xu Chen, Yankai Lin, Hao Sun, Ruihua Song, Xin Zhao, Jun Xu, Zhicheng Dou, Jun Wang, and Ji-Rong Wen. 2025b. User behavior simulation with large language model-based agents. ACM Trans. Inf. Syst., 43(2). find
  26. Xintao Wang, Heng Wang, Yifei Zhang, Xinfeng Yuan, Rui Xu, Jen-tse Huang, Siyu Yuan, Haoran Guo, Jiangjie Chen, Shuchang Zhou, Wei Wang, and Yanghua Xiao. 2025c. Coser: Coordinating llm-based persona simulation of established roles. In Forty-second International Conference on Machine Learning, ICML… find
  27. Tianjun Wei, Huizhong Guo, Yingpeng Du, Zhu Sun, Chen Huang, Dongxia Wang, and Jie Zhang. 2025. Mirroring users: Towards building preference-aligned user simulator with user feedback in recommendation. arXiv preprint arXiv:2508.18142.
  28. J. Williams. 2007. A method for evaluating and comparing user simulations: The cramér-von mises divergence. 2007 IEEE Workshop on Automatic Speech Recognition & Understanding (ASRU), pages 508– 513. find
  29. Ning Wu, Ming Gong, Linjun Shou, Shining Liang, and Daxin Jiang. 2023. Large language models are diverse role-players for summarization evaluation. In Natural Language Processing and Chinese Computing, pages 695–707, Cham. Springer Nature Switzerland.
  30. Shujin Wu, Yi R. Fung, Cheng Qian, Jeonghwan Kim, Dilek Hakkani-Tur, and Heng Ji. 2025. Aligning llms with individual preferences via interaction. In Proceedings of the 31st International Conference on Computational Linguistics, COLING 2025, Abu Dhabi, UAE, January 19-24, 2025, pages 7648–7662. Asso…
  31. Weiqi Wu, Hongqiu Wu, Lai Jiang, Xingyuan Liu, Hai Zhao, and Min Zhang. 2024. From role-play to drama-interaction: An LLM solution. In Findings of the Association for Computational Linguistics: ACL 2024, pages 3271–3290, Bangkok, Thailand. Association for Computational Linguistics. find
  32. Canwen Xu, Daya Guo, Nan Duan, and Julian J. McAuley. 2023. Baize: An open-source chat model with parameter-efficient tuning on self-chat data. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023, pages 6268– 6278. Ass…
  33. Yuxuan Yao, Han Wu, Qiling Xu, and Linqi Song. 2023. Fine-grained conversational decoding via isotropic and proximal search. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 58–70, Singapore. Association for Computational Linguistics. find
  34. Tong Zhang, Chen Huang, Yang Deng, Hongru Liang, Jia Liu, Zujie Wen, Wenqiang Lei, and Tat-Seng Chua. 2024. Strength lies in differences! improving strategy planning for non-collaborative dialogues via diversified user simulation. In Proceedings of the 2024 Conference on Empirical Methods in Natural… find
  35. Wenting Zhao, Xiang Ren, Jack Hessel, Claire Cardie, Yejin Choi, and Yuntian Deng. 2024. Wildchat: 1m chatGPT interaction logs in the wild. In The Twelfth International Conference on Learning Representations.
  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. 2024. Lmsys-chat-1m: A large-scale real-world llm conversation dataset. Preprint, arXiv:2309.11998.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wu-2026-thinking-alignment-scenario-oriented-user/