Elamur Research⌕
☰
home / papers / 2026 / Reasoning about Uncertainty: Do Reasoning Models Know When They Don't Know?

Reasoning about Uncertainty: Do Reasoning Models Know When They Don't Know?

Paper · 2026 · Zhiting Mei, Christina Zhang, Tenny Yin et al.

Summary

The paper asks whether reasoning LLMs can judge how reliable their own answers are. This matters for safe deployment, because these models often give confident but wrong answers. Top reasoning models are usually closed-source and do not expose logits. The authors therefore measure uncertainty through self-verbalized confidence, requesting only numeric scores to limit wording-related bias. They test Claude 3.7 Sonnet, o3-Mini, Gemini 2 Flash Thinking and DeepSeek R1 on several knowledge-heavy benchmarks. The study has three parts: overall calibration, the effect of more reasoning effort, and a new method called introspective uncertainty quantification (IUQ). In IUQ, a second, fresh session reviews the first session's chain of thought and answer, then assigns a confidence score.

The models tend to be overconfident, and this grows when they reason longer, especially once accuracy stops improving. Introspection lowers miscalibration for o3-Mini and DeepSeek R1, but it makes Claude 3.7 Sonnet worse. The authors conclude that new calibration-focused benchmarks and training methods are needed.

Key points

  • Models frequently report confidence above 85% (given in 5% steps) even on incorrect answers. Calibration is decent on knowledge-retrieval tasks but poor on expert-level domains, where high accuracy no longer hides overconfidence.
  • Prompt engineering improves calibration much less for reasoning models than it does for non-reasoning LLMs.
  • Increasing reasoning effort (low, medium, high) generally raises accuracy but also overconfidence, particularly on wrongly answered questions.
  • IUQ comes in three variants, IUQ-Low, IUQ-Medium and IUQ-High, evaluated on GPQA, StrategyQA and SimpleQA. Gemini was excluded because its reasoning traces could not be analysed adequately.
  • Averaged over models, ECE rises slightly with IUQ-Low and falls with IUQ-Medium and IUQ-High. The biggest gains appear on SimpleQA, while IUQ-High slightly worsens ECE on the easier StrategyQA.

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, 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 worksPhilosophy · 284 works

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

Cite

Cite as Zhiting Mei, Christina Zhang, Tenny Yin, Justin Lidard, Ola Sho and Anirudha Majumdar. 2026. Reasoning about Uncertainty: Do Reasoning Models Know When They Don't Know?. Findings of the Association for Computational Linguistics: EACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-eacl.178

@inproceedings{mei2026reasoning,
  title = {{Reasoning about Uncertainty: Do Reasoning Models Know When They Don't Know?}},
  author = {Zhiting Mei, Christina Zhang, Tenny Yin, Justin Lidard, Ola Sho and Anirudha Majumdar},
  booktitle = {Findings of the Association for Computational Linguistics: EACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-eacl.178},
  url = {https://doi.org/10.18653/v1/2026.findings-eacl.178}
}

References

59 references, 21 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Lukas Aichberger, Kajetan Schweighofer, Mykyta Ielanskyi, and Sepp Hochreiter. 2024. Semantically diverse language generation for uncertainty estimation in language models. arXiv preprint arXiv:2406.04306. — [arXiv:2406.04306](https://arxiv.org/abs/2406.04306) find
  2. Anthropic. 2024. Claude 3.7 sonnet system card. Technical report, Anthropic. find
  3. Art of Problem Solving. 2025. American Invitational Mathematics Examination. https://artofproblemsolving.com/wiki/index. php/American_Invitational_Mathematics_ Examination. ↗
  4. Amos Azaria and Tom Mitchell. 2023. The internal state of an llm knows when it’s lying. arXiv preprint arXiv:2304.13734.
  5. Zechen Bai, Pichao Wang, Tianjun Xiao, Tong He, Zongbo Han, Zheng Zhang, and Mike Zheng Shou. 2024. Hallucination of multimodal large language models: A survey. arXiv preprint arXiv:2404.18930.
  6. Yonatan Belinkov. 2022. Probing classifiers: Promises, shortcomings, and advances. Computational Linguistics, 48(1):207–219.
  7. Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Michal Podstawski, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Hubert Niewiadomski, Piotr Nyczyk, and 1 others. 2024. Graph of thoughts: Solving elaborate problems with large language models. In Proceedings of the AAAI Conference on A… find
  8. Peter Clark, Isaac Cowhey, Oren Etzioni, Tushar Khot, Ashish Sabharwal, Carissa Schoenick, and Oyvind Tafjord. 2018. Think you have solved question answering? try arc, the ai2 reasoning challenge. arXiv preprint arXiv:1803.05457.
  9. Jacob Dunefsky, Philippe Chlenski, and Neel Nanda. 2024. Transcoders find interpretable llm feature circuits. arXiv preprint arXiv:2406.11944. — [arXiv:2406.11944](https://arxiv.org/abs/2406.11944) find
  10. Ekaterina Fadeeva, Aleksandr Rubashevskii, Artem Shelmanov, Sergey Petrakov, Haonan Li, Hamdy Mubarak, Evgenii Tsymbalov, Gleb Kuzmin, Alexander Panchenko, Timothy Baldwin, and 1 others. 2024. Fact-checking the output of large language models via token-level uncertainty quantification. arXiv preprin…
  11. Javier Ferrando, Oscar Obeso, Senthooran Rajamanoharan, and Neel Nanda. 2024. Do i know this entity? knowledge awareness and hallucinations in language models. arXiv preprint arXiv:2411.14257. — [arXiv:2411.14257](https://arxiv.org/abs/2411.14257) find
  12. Yao Fu, Hao Peng, Ashish Sabharwal, Peter Clark, and Tushar Khot. 2022. Complexity-based prompting for multi-step reasoning. arXiv preprint arXiv:2210.00720.
  13. Leo Gao, Tom Dupré la Tour, Henk Tillman, Gabriel Goh, Rajan Troll, Alec Radford, Ilya Sutskever, Jan Leike, and Jeffrey Wu. 2024. Scaling and evaluating sparse autoencoders. arXiv preprint arXiv:2406.04093. find
  14. Mor Geva, Daniel Khashabi, Elad Segal, Tushar Khot, Dan Roth, and Jonathan Berant. 2021. Did aristotle use a laptop? a question answering benchmark with implicit reasoning strategies. Transactions of the Association for Computational Linguistics, 9:346– 361. find
  15. Mor Geva, Roei Schuster, Jonathan Berant, and Omer Levy. 2020. Transformer feed-forward layers are key-value memories. arXiv preprint arXiv:2012.14913. find
  16. Tilmann Gneiting and Adrian E Raftery. 2007. Strictly proper scoring rules, prediction, and estimation. Journal of the American statistical Association, 102(477):359–378. find
  17. Chuan Guo, Geoff Pleiss, Yu Sun, and Kilian Q. Weinberger. 2017. On calibration of modern neural networks. In Proceedings of the 34th International Conference on Machine Learning, volume 70 of Proceedings of Machine Learning Research, pages 1321– 1330. PMLR. find
  18. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, and 1 others. 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948.
  19. Shibo Hao, Yi Gu, Haotian Luo, Tianyang Liu, Xiyan Shao, Xinyuan Wang, Shuhua Xie, Haodi Ma, Adithya Samavedhi, Qiyue Gao, and 1 others. 2024. Llm reasoners: New evaluation, library, and analysis of step-by-step reasoning with large language models. arXiv preprint arXiv:2404.05221. — [arXiv:2404.05221](https://arxiv.org/abs/2404.05221) find
  20. Shibo Hao, Yi Gu, Haodi Ma, Joshua Jiahua Hong, Zhen Wang, Daisy Zhe Wang, and Zhiting Hu. 2023. Reasoning with language model is planning with world model. arXiv preprint arXiv:2305.14992. find
  21. Stephanie Lin, Jacob Hilton, and Owain Evans. 2022. Teaching models to express their uncertainty in words. arXiv preprint arXiv:2205.14334. find
  22. Chen Ling, Xujiang Zhao, Wei Cheng, Yanchi Liu, Yiyou Sun, Xuchao Zhang, Mika Oishi, Takao Osaki, Katsushi Matsuda, Jie Ji, and 1 others. 2024. Uncertainty decomposition and quantification for in-context learning of large language models. arXiv e-prints, pages arXiv–2402. find
  23. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2020. Measuring massive multitask language understanding. arXiv preprint arXiv:2009.03300.
  24. Hanchao Liu, Wenyuan Xue, Yifei Chen, Dapeng Chen, Xiutian Zhao, Ke Wang, Liping Hou, Rongjun Li, and Wei Peng. 2024. A survey on hallucination in large vision-language models. arXiv preprint arXiv:2402.00253. find
  25. Tanya Hewitt, Samia Chreim, and Alan Forster. 2016. Double checking: a second look. Journal of evaluation in clinical practice, 22(2):267–274. find
  26. 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
  27. Jie Huang and Kevin Chen-Chuan Chang. 2022. Towards reasoning in large language models: A survey. arXiv preprint arXiv:2212.10403.
  28. 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.
  29. Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Helyar, Aleksander Madry, Alex Beutel, Alex Carney, and 1 others. 2024. Openai o1 system card. arXiv preprint arXiv:2412.16720.
  30. Laurent Valentin Jospin, Hamid Laga, Farid Boussaid, Wray Buntine, and Mohammed Bennamoun. 2022. Hands-on bayesian neural networks—a tutorial for deep learning users. IEEE Computational Intelligence Magazine, 17(2):29–48. find
  31. Saurav Kadavath, Tom Conerly, Amanda Askell, Tom Henighan, Dawn Drain, Ethan Perez, Nicholas Schiefer, Zac Hatfield-Dodds, Nova DasSarma, Eli Tran-Johnson, and 1 others. 2022. Language models (mostly) know what they know. arXiv preprint arXiv:2207.05221. find
  32. Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. 2022. Large language models are zero-shot reasoners. Advances in neural information processing systems, 35:22199– 22213.
  33. Sabrina J Mielke, Arthur Szlam, Emily Dinan, and Y-Lan Boureau. 2022. Reducing conversational agents’ overconfidence through linguistic calibration. Transactions of the Association for Computational Linguistics, 10:857–872. find
  34. Mikhail Mirzayanov and Codeforces Team. 2025. Codeforces. https://codeforces.com. Competitive Programming Platform. ↗
  35. Mahdi P. Naeini, Gregory Cooper, and Milos Hauskrecht. 2015. Obtaining well calibrated probabilities using bayesian binning. Proceedings of the AAAI Conference on Artificial Intelligence, 29. find
  36. OpenAI. 2024a. Gpt-4 technical report. Preprint, arXiv:2303.08774.
  37. OpenAI. 2024b. Openai o3-mini system card. Technical report, OpenAI. find
  38. David Rein, Betty Li Hou, Asa Cooper Stickland, Jackson Petty, Richard Yuanzhe Pang, Julien Dirani, Julian Michael, and Samuel R Bowman. 2023. Gpqa: A graduate-level google-proof q&a benchmark. arXiv preprint arXiv:2311.12022.
  39. Ola Shorinwa, Zhiting Mei, Justin Lidard, Allen Z. Ren, and Anirudha Majumdar. 2025. A survey on uncertainty quantification of large language models: Taxonomy, open research challenges, and future directions. ACM Comput. Surv. find
  40. Jody M Shynkaruk and Valerie A Thompson. 2006. Confidence and accuracy in deductive reasoning. Memory & cognition, 34(3):619–632. find
  41. Lorenz Kuhn, Yarin Gal, and Sebastian Farquhar. 2023. Semantic uncertainty: Linguistic invariances for uncertainty estimation in natural language generation. arXiv preprint arXiv:2302.09664. find
  42. Eric Michael Smith, Diana Gonzalez-Rico, Emily Dinan, and Y-Lan Boureau. 2020. Controlling style in generated dialogue. arXiv preprint arXiv:2009.10855. — [arXiv:2009.10855](https://arxiv.org/abs/2009.10855) find
  43. Kaiqu Liang, Zixu Zhang, and Jaime F Fisac. 2024. Introspective planning: Aligning robots’ uncertainty with inherent task ambiguity. Advances in Neural Information Processing Systems, 37:71998–72031. find
  44. Alex Tamkin, Mohammad Taufeeque, and Noah D Goodman. 2023. Codebook features: Sparse and discrete interpretability for neural networks. arXiv preprint arXiv:2310.17230. — [arXiv:2310.17230](https://arxiv.org/abs/2310.17230) find
  45. Sree Harsha Tanneru, Chirag Agarwal, and Himabindu Lakkaraju. 2024. Quantifying uncertainty in natural language explanations of large language models. In International Conference on Artificial Intelligence and Statistics, pages 1072–1080. PMLR. find
  46. Zeyu Yun, Yubei Chen, Bruno A Olshausen, and Yann LeCun. 2021. Transformer visualization via dictionary learning: contextualized embedding as a linear superposition of transformer factors. arXiv preprint arXiv:2103.15949. find
  47. Gemini Team, Petko Georgiev, Ving Ian Lei, Ryan Burnell, Libin Bai, Anmol Gulati, Garrett Tanzer, Damien Vincent, Zhufeng Pan, Shibo Wang, and 1 others. 2024. Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context. arXiv preprint arXiv:2403.05530.
  48. Hanning Zhang, Shizhe Diao, Yong Lin, Yi Fung, Qing Lian, Xingyao Wang, Yangyi Chen, Heng Ji, and Tong Zhang. 2024. R-tuning: Instructing large language models to say ‘i don’t know’. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics:… find
  49. Qwen Team. 2025. Qwq-32b: Embracing the power of reinforcement learning. find
  50. Katherine Tian, Eric Mitchell, Allan Zhou, Archit Sharma, Rafael Rafailov, Huaxiu Yao, Chelsea Finn, and Christopher D Manning. 2023. Just ask for calibration: Strategies for eliciting calibrated confidence scores from language models fine-tuned with human feedback. arXiv preprint arXiv:2305.14975. find
  51. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2022. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171.
  52. Zhuosheng Zhang, Aston Zhang, Mu Li, and Alex Smola. 2022. Automatic chain of thought prompting in large language models. arXiv preprint arXiv:2210.03493.
  53. Denny Zhou, Nathanael Schärli, Le Hou, Jason Wei, Nathan Scales, Xuezhi Wang, Dale Schuurmans, Claire Cui, Olivier Bousquet, Quoc Le, and 1 others. 2022. Least-to-most prompting enables complex reasoning in large language models. arXiv preprint arXiv:2205.10625.
  54. Jason Wei, Nguyen Karina, Hyung Won Chung, Yunxin Joy Jiao, Spencer Papay, Amelia Glaese, John Schulman, and William Fedus. 2024. Measuring short-form factuality in large language models. arXiv preprint arXiv:2411.04368.
  55. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824– 24837.
  56. Zhaofeng Wu, Linlu Qiu, Alexis Ross, Ekin Akyürek, Boyuan Chen, Bailin Wang, Najoung Kim, Jacob Andreas, and Yoon Kim. 2024. Reasoning or reciting? exploring the capabilities and limitations of language models through counterfactual tasks. In Proceedings of the 2024 Conference of the North American… find
  57. Yijun Xiao and William Yang Wang. 2021. On hallucination and predictive uncertainty in conditional language generation. arXiv preprint arXiv:2103.15025.
  58. Miao Xiong, Zhiyuan Hu, Xinyang Lu, Yifei Li, Jie Fu, Junxian He, and Bryan Hooi. 2023. Can llms express their uncertainty? an empirical evaluation of confidence elicitation in llms. arXiv preprint arXiv:2306.13063. find
  59. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2023. Tree of thoughts: Deliberate problem solving with large language models. Advances in neural information processing systems, 36:11809–11822. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/mei-2026-reasoning-about-uncertainty-do-reasoning/