---
title: "Balancing Classification and Calibration Performance in"
description: "The base model also offers almost no calibrated reasoning paths for RL to reinforce. 47 references."
authors: ["Duygu Nur Yaldiz", "Evangelia Spiliopoulou", "Zheng Qi", "Siddharth Varia", "Srikanth Doss", "Nikolaos Pappas"]
year: 2026
venue: "Findings of the Association for Computational Linguistics: ACL 2026"
source: "https://aclanthology.org/2026.findings-acl.610/"
doi: "10.18653/v1/2026.findings-acl.610"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/yaldiz-2026-balancing-classification-calibration-performance-decision/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/yaldiz-2026-balancing-classification-calibration-performance-decision/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/24/) / Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning

# Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning

**Authors:** Duygu Nur Yaldiz, [Evangelia Spiliopoulou](https://md.research.elamur.ai/free/en/authors/evangelia-spiliopoulou/), [Zheng Qi](https://md.research.elamur.ai/free/en/authors/zheng-qi/), [Siddharth Varia](https://md.research.elamur.ai/free/en/authors/siddharth-varia/), [Srikanth Doss](https://md.research.elamur.ai/free/en/authors/srikanth-doss/) and [Nikolaos Pappas](https://md.research.elamur.ai/free/en/authors/nikolaos-pappas/) · **Year:** 2026 · **Venue:** [Findings of the Association for Computational Linguistics: ACL 2026](https://md.research.elamur.ai/free/en/venues/findings/) ([Findings of the Association for Computational Linguistics 2026](https://md.research.elamur.ai/free/en/venues/findings/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.findings-acl.610/) · [PDF](https://aclanthology.org/2026.findings-acl.610.pdf) · [DOI 10.18653/v1/2026.findings-acl.610](https://doi.org/10.18653/v1/2026.findings-acl.610) · [OpenAlex](https://openalex.org/W7166847074)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
LLMs used for decisions such as content moderation or clinical support need confidence scores that tell downstream systems when to fall back to a larger model or a human. The authors study how two common fine-tuning paradigms affect probability-based confidence, defined as the probability of the final decision token. They evaluate it with ECE, reliability diagrams and confidence histograms. Across several decision-making benchmarks, they find a consistent trade-off. Reinforcement learning with verifiable rewards (RLVR) gives larger accuracy gains but makes models extremely overconfident. Supervised fine-tuning (SFT) keeps confidence closer to accuracy, even under distribution shift, but improves accuracy less. Diagnostic experiments show two causes. After reasoning, the decision token only reads off the conclusion and carries no uncertainty. The base model also offers almost no calibrated reasoning paths for RL to reinforce. The authors therefore modify GRPO so that training directly shapes decision-token probabilities. This calibration-aware RL keeps RLVR-level accuracy while lowering ECE by up to 9 points, which makes cheap single-pass confidence usable in low-latency systems.

## Key points
- Confidence is the decision token's probability, P(y_d | x, y_\<d). Multi-token answers are recast as atomic options (A, B, C, D), so every decision is a single token.
- ECE is computed with equal-size bins, a variant Nixon et al. (2019) call Adaptive Calibration Error. It is read together with reliability diagrams and confidence histograms.
- The diagnosis has two parts: decision tokens inherit overconfidence from the reasoning trace, and most base-model paths are already overconfident.
- The proposed method changes GRPO to calibrate the decision token while keeping reasoning and decision consistent. It reduces ECE by up to 9 points, matches RLVR accuracy and generalizes to out-of-distribution data.
- Experiments use only the Qwen3 family up to 8B parameters. Applying post-hoc calibration (e.g., temperature scaling) on top of the method gave the best overall calibration.

*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 Southern California](https://md.research.elamur.ai/free/en/universities/university-of-southern-california/) — University, Los Angeles, United States · 175 works · 37,047 citations
- [Amazon](https://md.research.elamur.ai/free/en/companies/amazon/) — Company, Seattle, United States · 128 works · 41,697 citations
- [Oracle](https://md.research.elamur.ai/free/en/companies/oracle/) — Company, Austin, United States · 9 works · 1,235 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works

🧭 Topics (1): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,894 works

## Cite

Cite as: Duygu Nur Yaldiz, Evangelia Spiliopoulou, Zheng Qi, Siddharth Varia, Srikanth Doss and Nikolaos Pappas. 2026. Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{yaldiz2026balancing,
  title = {{Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning}},
  author = {Duygu Nur Yaldiz, Evangelia Spiliopoulou, Zheng Qi, Siddharth Varia, Srikanth Doss and Nikolaos Pappas},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.610},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.610}
}
```

- [← previous Baflah-lamri at NAKBA-NLP 2026: Manual Ground Truth Enrichment](https://md.research.elamur.ai/free/en/papers/abdelouahab-2026-baflah-lamri-nakba-nlp-2026/)
- [next → Balancing FAIR and GDPR: A Governance Framework for Oral Archives](https://md.research.elamur.ai/free/en/papers/mercatanti-2026-balancing-fair-gdpr-governance-framework/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/24/)

## References

47 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Yavuz Bakman, Sungmin Kang, Zhiqi Huang, Duygu Nur Yaldiz, Catarina G. Belém, Chenyang Zhu, Anoop Kumar, Alfy Samuel, Salman Avestimehr, Daben Liu, and Sai Praneeth Karimireddy. 2025a. Uncertainty as feature gaps: Epistemic uncertainty quantification of llms in contextual question-answering. Preprin… — [arXiv:2510.02671](https://arxiv.org/abs/2510.02671) [find](https://scholar.google.com/scholar?q=Yavuz+Bakman%2C+Sungmin+Kang%2C+Zhiqi+Huang%2C+Duygu+Nur+Yaldiz%2C+Catarina+G%2E+Bel%C3%A9m%2C+Chenyang+Zhu%2C+Anoop+Kumar%2C+Alfy+Samuel%2C+Salman+Avestimehr%2C+Daben+Liu%2C+and+Sai+Praneeth+Karimireddy%2E+2025a%2E+Uncertainty+as+)
2. Yavuz Faruk Bakman, Duygu Nur Yaldiz, Baturalp Buyukates, Chenyang Tao, Dimitrios Dimitriadis, and Salman Avestimehr. 2024. MARS: Meaning-aware response scoring for uncertainty estimation in generative LLMs. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (… [find](https://scholar.google.com/scholar?q=Yavuz+Faruk+Bakman%2C+Duygu+Nur+Yaldiz%2C+Baturalp+Buyukates%2C+Chenyang+Tao%2C+Dimitrios+Dimitriadis%2C+and+Salman+Avestimehr%2E+2024%2E+MARS%3A+Meaning%2Daware+response+scoring+for+uncertainty+estimation+in+generativ)
3. Yavuz Faruk Bakman, Duygu Nur Yaldiz, Sungmin Kang, Tuo Zhang, Baturalp Buyukates, Salman Avestimehr, and Sai Praneeth Karimireddy. 2025b. Reconsidering LLM uncertainty estimation methods in the wild. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume… [find](https://scholar.google.com/scholar?q=Yavuz+Faruk+Bakman%2C+Duygu+Nur+Yaldiz%2C+Sungmin+Kang%2C+Tuo+Zhang%2C+Baturalp+Buyukates%2C+Salman+Avestimehr%2C+and+Sai+Praneeth+Karimireddy%2E+2025b%2E+Reconsidering+LLM+uncertainty+estimation+methods+in+the+wild%2E)
4. Neeloy Chakraborty, Melkior Ornik, and Katherine Rose Driggs-Campbell. 2025. Hallucination detection in foundation models for decision-making: A flexible definition and review of the state of the art. ACM Comput. Surv., 57(7):188:1–188:35. [find](https://scholar.google.com/scholar?q=Neeloy+Chakraborty%2C+Melkior+Ornik%2C+and+Katherine+Rose+Driggs%2DCampbell%2E+2025%2E+Hallucination+detection+in+foundation+models+for+decision%2Dmaking%3A+A+flexible+definition+and+review+of+the+state+of+the+art%2E)
5. Mehul Damani, Isha Puri, Stewart Slocum, Idan Shenfeld, Leshem Choshen, Yoon Kim, and Jacob Andreas. 2025. Beyond binary rewards: Training lms to reason about their uncertainty. Preprint, arXiv:2507.16806. [find](https://scholar.google.com/scholar?q=Mehul+Damani%2C+Isha+Puri%2C+Stewart+Slocum%2C+Idan+Shenfeld%2C+Leshem+Choshen%2C+Yoon+Kim%2C+and+Jacob+Andreas%2E+2025%2E+Beyond+binary+rewards%3A+Training+lms+to+reason+about+their+uncertainty%2E+Preprint%2C+arXiv%3A2507%2E1)
6. DeepSeek-AI, Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, Xiaokang Zhang, Xingkai Yu, Yu Wu, Z. F. Wu, Zhibin Gou, Zhihong Shao, Zhuoshu Li, Ziyi Gao, and 181 others. 2025. [Deepseek-r1: Incentivizing reasoning capability in ll…](https://md.research.elamur.ai/free/en/papers/deepseek-ai-2025-deepseek-r1-incentivizing-reasoning-capability/)
7. Jinhao Duan, Hao Cheng, Shiqi Wang, Alex Zavalny, Chenan Wang, Renjing Xu, Bhavya Kailkhura, and Kaidi Xu. 2024. Shifting attention to relevance: Towards the predictive uncertainty quantification of free-form large language models. In Proceedings of the 62nd Annual Meeting of the Association for Com… [find](https://scholar.google.com/scholar?q=Jinhao+Duan%2C+Hao+Cheng%2C+Shiqi+Wang%2C+Alex+Zavalny%2C+Chenan+Wang%2C+Renjing+Xu%2C+Bhavya+Kailkhura%2C+and+Kaidi+Xu%2E+2024%2E+Shifting+attention+to+relevance%3A+Towards+the+predictive+uncertainty+quantification+of+f)
8. Eva Eigner and Thorsten Händler. 2024. Determinants of llm-assisted decision-making. Preprint, arXiv:2402.17385. — [arXiv:2402.17385](https://arxiv.org/abs/2402.17385) [find](https://scholar.google.com/scholar?q=Eva+Eigner+and+Thorsten+H%C3%A4ndler%2E+2024%2E+Determinants+of+llm%2Dassisted+decision%2Dmaking%2E+Preprint%2C+arXiv%3A2402%2E17385%2E+%E2%80%94+%5BarXiv%3A2402%2E17385%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2402%2E17385%29)
9. 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, ICML’17, page 1321–1330. [find](https://scholar.google.com/scholar?q=Chuan+Guo%2C+Geoff+Pleiss%2C+Yu+Sun%2C+and+Kilian+Q%2E+Weinberger%2E+2017%2E+On+calibration+of+modern+neural+networks%2E+In+Proceedings+of+the+34th+International+Conference+on+Machine+Learning+%2D+Volume+70%2C+ICML%E2%80%9917%2C)
10. Seungju Han, Kavel Rao, Allyson Ettinger, Liwei Jiang, Bill Yuchen Lin, Nathan Lambert, Yejin Choi, and Nouha Dziri. 2024. [Wildguard: Open one-stop moderation tools for safety risks, jailbreaks, and refusals of llms](https://md.research.elamur.ai/free/en/papers/han-2024-wildguard-open-one-stop-moderation/). In Advances in Neural Information Processing Systems, volume 37, pages 8093–8131. C…
11. Edward J Hu, yelong shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. [LoRA: Low-rank adaptation of large language models](https://md.research.elamur.ai/free/en/papers/hu-2021-lora-low-rank-adaptation-large/). In International Conference on Learning Representations.
12. Jingcheng Hu, Yinmin Zhang, Qi Han, Daxin Jiang, Xiangyu Zhang, and Heung-Yeung Shum. 2025. [Open-reasoner-zero: An open source approach to scaling up reinforcement learning on the base model](https://md.research.elamur.ai/free/en/papers/hu-2025-open-reasoner-zero-open-source/). Preprint, arXiv:2503.24290. 12546 10
13. Jerry Huang, Peng Lu, and QIUHAO Zeng. 2025. Calibrated language models and how to find them with label smoothing. In Forty-second International Conference on Machine Learning. [find](https://scholar.google.com/scholar?q=Jerry+Huang%2C+Peng+Lu%2C+and+QIUHAO+Zeng%2E+2025%2E+Calibrated+language+models+and+how+to+find+them+with+label+smoothing%2E+In+Forty%2Dsecond+International+Conference+on+Machine+Learning%2E)
14. Hakan Inan, Kartikeya Upasani, Jianfeng Chi, Rashi Rungta, Krithika Iyer, Yuning Mao, Michael Tontchev, Qing Hu, Brian Fuller, Davide Testuggine, and Madian Khabsa. 2023. [Llama guard: Llm-based input-output safeguard for human-ai conversations](https://md.research.elamur.ai/free/en/papers/inan-2023-llama-guard-llm-based-input/). Preprint, arXiv:2312.06674.
15. Saurav Kadavath, Tom Conerly, Amanda Askell, Tom Henighan, Dawn Drain, Ethan Perez, Nicholas Schiefer, Zac Hatfield-Dodds, Nova DasSarma, Eli Tran-Johnson, Scott Johnston, Sheer El-Showk, Andy Jones, Nelson Elhage, Tristan Hume, Anna Chen, Yuntao Bai, Sam Bowman, Stanislav Fort, and 17 others. 2022.… [find](https://scholar.google.com/scholar?q=Saurav+Kadavath%2C+Tom+Conerly%2C+Amanda+Askell%2C+Tom+Henighan%2C+Dawn+Drain%2C+Ethan+Perez%2C+Nicholas+Schiefer%2C+Zac+Hatfield%2DDodds%2C+Nova+DasSarma%2C+Eli+Tran%2DJohnson%2C+Scott+Johnston%2C+Sheer+El%2DShowk%2C+Andy+Jones%2C+)
16. Lorenz Kuhn, Yarin Gal, and Sebastian Farquhar. 2023. [Semantic uncertainty: Linguistic invariances for uncertainty estimation in natural language generation](https://md.research.elamur.ai/free/en/papers/kuhn-2023-semantic-uncertainty-linguistic-invariances-uncertainty/). In The Eleventh International Conference on Learning Representations.
17. Todor Markov, Chong Zhang, Sandhini Agarwal, Tyna Eloundou, Teddy Lee, Steven Adler, Angela Jiang, and Lilian Weng. 2023. [A holistic approach to undesired content detection in the real world](https://md.research.elamur.ai/free/en/papers/markov-2022-holistic-approach-undesired-content-detection/). Preprint, arXiv:2208.03274.
18. Todor Mihaylov, Peter Clark, Tushar Khot, and Ashish Sabharwal. 2018. Can a suit of armor conduct electricity? a new dataset for open book question answering. In Conference on Empirical Methods in Natural Language Processing. [find](https://scholar.google.com/scholar?q=Todor+Mihaylov%2C+Peter+Clark%2C+Tushar+Khot%2C+and+Ashish+Sabharwal%2E+2018%2E+Can+a+suit+of+armor+conduct+electricity%3F+a+new+dataset+for+open+book+question+answering%2E+In+Conference+on+Empirical+Methods+in+Nat)
19. Alexander V Nikitin, Jannik Kossen, Yarin Gal, and Pekka Marttinen. 2024. Kernel language entropy: Fine-grained uncertainty quantification for LLMs from semantic similarities. In The Thirty-eighth Annual Conference on Neural Information Processing Systems. [find](https://scholar.google.com/scholar?q=Alexander+V+Nikitin%2C+Jannik+Kossen%2C+Yarin+Gal%2C+and+Pekka+Marttinen%2E+2024%2E+Kernel+language+entropy%3A+Fine%2Dgrained+uncertainty+quantification+for+LLMs+from+semantic+similarities%2E+In+The+Thirty%2Deighth+Ann)
20. Jeremy Nixon, Michael W. Dusenberry, Linchuan Zhang, Ghassen Jerfel, and Dustin Tran. 2019. Measuring calibration in deep learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops. [find](https://scholar.google.com/scholar?q=Jeremy+Nixon%2C+Michael+W%2E+Dusenberry%2C+Linchuan+Zhang%2C+Ghassen+Jerfel%2C+and+Dustin+Tran%2E+2019%2E+Measuring+calibration+in+deep+learning%2E+In+Proceedings+of+the+IEEE%2FCVF+Conference+on+Computer+Vision+and+Pat)
21. OpenAI. 2023. [GPT-4 Technical Report](https://md.research.elamur.ai/free/en/papers/openai-2023-gpt-4-technical-report/). Preprint, arXiv:2303.08774.
22. Ananya Kumar, Percy S Liang, and Tengyu Ma. 2019. Verified uncertainty calibration. In Advances in Neural Information Processing Systems, volume 32. Curran Associates, Inc. [find](https://scholar.google.com/scholar?q=Ananya+Kumar%2C+Percy+S+Liang%2C+and+Tengyu+Ma%2E+2019%2E+Verified+uncertainty+calibration%2E+In+Advances+in+Neural+Information+Processing+Systems%2C+volume+32%2E+Curran+Associates%2C+Inc%2E)
23. John Platt and 1 others. 1999. Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods. Advances in large margin classifiers, 10(3):61–74. [find](https://scholar.google.com/scholar?q=John+Platt+and+1+others%2E+1999%2E+Probabilistic+outputs+for+support+vector+machines+and+comparisons+to+regularized+likelihood+methods%2E+Advances+in+large+margin+classifiers%2C+10%283%29%3A61%E2%80%9374%2E)
24. Jixuan Leng, Chengsong Huang, Banghua Zhu, and Jiaxin Huang. 2025. Taming overconfidence in LLMs: Reward calibration in RLHF. In The Thirteenth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Jixuan+Leng%2C+Chengsong+Huang%2C+Banghua+Zhu%2C+and+Jiaxin+Huang%2E+2025%2E+Taming+overconfidence+in+LLMs%3A+Reward+calibration+in+RLHF%2E+In+The+Thirteenth+International+Conference+on+Learning+Representations%2E)
25. Paul Röttger, Hannah Rose Kirk, Bertie Vidgen, Giuseppe Attanasio, Federico Bianchi, and Dirk Hovy. 2024. [Xstest: A test suite for identifying exaggerated safety behaviours in large language models](https://md.research.elamur.ai/free/en/papers/rottger-2023-xstest-test-suite-identifying-exaggerated/). Preprint, arXiv:2308.01263.
26. Stephanie Lin, Jacob Hilton, and Owain Evans. 2022. Teaching models to express their uncertainty in words. Transactions on Machine Learning Research. [find](https://scholar.google.com/scholar?q=Stephanie+Lin%2C+Jacob+Hilton%2C+and+Owain+Evans%2E+2022%2E+Teaching+models+to+express+their+uncertainty+in+words%2E+Transactions+on+Machine+Learning+Research%2E)
27. Zhen Lin, Shubhendu Trivedi, and Jimeng Sun. 2024. Generating with confidence: Uncertainty quantification for black-box large language models. Transactions on Machine Learning Research. [find](https://scholar.google.com/scholar?q=Zhen+Lin%2C+Shubhendu+Trivedi%2C+and+Jimeng+Sun%2E+2024%2E+Generating+with+confidence%3A+Uncertainty+quantification+for+black%2Dbox+large+language+models%2E+Transactions+on+Machine+Learning+Research%2E)
28. Hongfu Liu, Hengguan Huang, Xiangming Gu, Hao Wang, and Ye Wang. 2025. On calibration of LLM-based guard models for reliable content moderation. In The Thirteenth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Hongfu+Liu%2C+Hengguan+Huang%2C+Xiangming+Gu%2C+Hao+Wang%2C+and+Ye+Wang%2E+2025%2E+On+calibration+of+LLM%2Dbased+guard+models+for+reliable+content+moderation%2E+In+The+Thirteenth+International+Conference+on+Learning+)
29. Shudong Liu, Zhaocong Li, Xuebo Liu, Runzhe Zhan, Derek F. Wong, Lidia S. Chao, and Min Zhang. 2024. Can LLMs learn uncertainty on their own? expressing uncertainty effectively in a self-training manner. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages… [find](https://scholar.google.com/scholar?q=Shudong+Liu%2C+Zhaocong+Li%2C+Xuebo+Liu%2C+Runzhe+Zhan%2C+Derek+F%2E+Wong%2C+Lidia+S%2E+Chao%2C+and+Min+Zhang%2E+2024%2E+Can+LLMs+learn+uncertainty+on+their+own%3F+expressing+uncertainty+effectively+in+a+self%2Dtraining+mann)
30. Andrey Malinin and Mark Gales. 2021. Uncertainty estimation in autoregressive structured prediction. In International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Andrey+Malinin+and+Mark+Gales%2E+2021%2E+Uncertainty+estimation+in+autoregressive+structured+prediction%2E+In+International+Conference+on+Learning+Representations%2E)
31. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y. K. Li, Y. Wu, and Daya Guo. 2024. [Deepseekmath: Pushing the limits of mathematical reasoning in open language models](https://md.research.elamur.ai/free/en/papers/shao-2024-deepseekmath-pushing-limits-mathematical-reasoning/). Preprint, arXiv:2402.03300.
32. Paul Stangel, David Bani-Harouni, Chantal Pellegrini, Ege Özsoy, Kamilia Zaripova, Matthias Keicher, and Nassir Navab. 2025. Rewarding doubt: A reinforcement learning approach to calibrated confidence expression of large language models. Preprint, arXiv:2503.02623. — [arXiv:2503.02623](https://arxiv.org/abs/2503.02623) [find](https://scholar.google.com/scholar?q=Paul+Stangel%2C+David+Bani%2DHarouni%2C+Chantal+Pellegrini%2C+Ege+%C3%96zsoy%2C+Kamilia+Zaripova%2C+Matthias+Keicher%2C+and+Nassir+Navab%2E+2025%2E+Rewarding+doubt%3A+A+reinforcement+learning+approach+to+calibrated+confidence)
33. Elias Stengel-Eskin, Peter Hase, and Mohit Bansal. 2024. Lacie: Listener-aware finetuning for calibration in large language models. In Advances in Neural Information Processing Systems, volume 37, pages 43080–43106. Curran Associates, Inc. [find](https://scholar.google.com/scholar?q=Elias+Stengel%2DEskin%2C+Peter+Hase%2C+and+Mohit+Bansal%2E+2024%2E+Lacie%3A+Listener%2Daware+finetuning+for+calibration+in+large+language+models%2E+In+Advances+in+Neural+Information+Processing+Systems%2C+volume+37%2C+pag)
34. Alon Talmor, Jonathan Herzig, Nicholas Lourie, and Jonathan Berant. 2019. [CommonsenseQA: A question answering challenge targeting commonsense knowledge](https://md.research.elamur.ai/free/en/papers/talmor-2019-commonsenseqa-question-answering-challenge-targeting/). In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,…
35. Shuchang Tao, Liuyi Yao, Hanxing Ding, Yuexiang Xie, Qi Cao, Fei Sun, Jinyang Gao, Huawei Shen, and Bolin Ding. 2024. When to trust LLMs: Aligning confidence with response quality. In Findings of the Association for Computational Linguistics: ACL 2024, pages 5984–5996, Bangkok, Thailand. Association… [find](https://scholar.google.com/scholar?q=Shuchang+Tao%2C+Liuyi+Yao%2C+Hanxing+Ding%2C+Yuexiang+Xie%2C+Qi+Cao%2C+Fei+Sun%2C+Jinyang+Gao%2C+Huawei+Shen%2C+and+Bolin+Ding%2E+2024%2E+When+to+trust+LLMs%3A+Aligning+confidence+with+response+quality%2E+In+Findings+of+the+)
36. Katherine Tian, Eric Mitchell, Allan Zhou, Archit Sharma, Rafael Rafailov, Huaxiu Yao, Chelsea Finn, and Christopher Manning. 2023. Just ask for calibration: Strategies for eliciting calibrated confidence scores from language models fine-tuned with human feedback. In Proceedings of the 2023 Conferen… [find](https://scholar.google.com/scholar?q=Katherine+Tian%2C+Eric+Mitchell%2C+Allan+Zhou%2C+Archit+Sharma%2C+Rafael+Rafailov%2C+Huaxiu+Yao%2C+Chelsea+Finn%2C+and+Christopher+Manning%2E+2023%2E+Just+ask+for+calibration%3A+Strategies+for+eliciting+calibrated+confid)
37. 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](https://github.com/huggingface/trl). [↗](https://github.com/huggingface/trl)
38. Boshi Wang, Sewon Min, Xiang Deng, Jiaming Shen, You Wu, Luke Zettlemoyer, and Huan Sun. 2023. [Towards understanding chain-of-thought prompting: An empirical study of what matters](https://md.research.elamur.ai/free/en/papers/wang-2023-understanding-chain-thought-prompting-empirical/). In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pag…
39. Duygu Nur Yaldiz, Yavuz Faruk Bakman, Baturalp Buyukates, Chenyang Tao, Anil Ramakrishna, Dimitrios Dimitriadis, Jieyu Zhao, and Salman Avestimehr. 2025. Do not design, learn: A trainable scoring function for uncertainty estimation in generative LLMs. In Findings of the Association for Computational… [find](https://scholar.google.com/scholar?q=Duygu+Nur+Yaldiz%2C+Yavuz+Faruk+Bakman%2C+Baturalp+Buyukates%2C+Chenyang+Tao%2C+Anil+Ramakrishna%2C+Dimitrios+Dimitriadis%2C+Jieyu+Zhao%2C+and+Salman+Avestimehr%2E+2025%2E+Do+not+design%2C+learn%3A+A+trainable+scoring+func)
40. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, and 41 others. 2025. [Qwen3 technical report](https://md.research.elamur.ai/free/en/papers/yang-2025-qwen3-technical-report/). Preprint, arXiv:2505.09388.
41. Bianca Zadrozny and Charles Elkan. 2002. Transforming classifier scores into accurate multiclass probability estimates. In Proceedings of the Eighth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’02, page 694–699, New York, NY, USA. Association for Computing Machine… [find](https://scholar.google.com/scholar?q=Bianca+Zadrozny+and+Charles+Elkan%2E+2002%2E+Transforming+classifier+scores+into+accurate+multiclass+probability+estimates%2E+In+Proceedings+of+the+Eighth+ACM+SIGKDD+International+Conference+on+Knowledge+Di)
42. Han Zhou, Xingchen Wan, Lev Proleev, Diana Mincu, Jilin Chen, Katherine A Heller, and Subhrajit Roy. 2024. Batch calibration: Rethinking calibration for in-context learning and prompt engineering. In The Twelfth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Han+Zhou%2C+Xingchen+Wan%2C+Lev+Proleev%2C+Diana+Mincu%2C+Jilin+Chen%2C+Katherine+A+Heller%2C+and+Subhrajit+Roy%2E+2024%2E+Batch+calibration%3A+Rethinking+calibration+for+in%2Dcontext+learning+and+prompt+engineering%2E+In+)
43. Cheng Wang, Gyuri Szarvas, Georges Balazs, Pavel Danchenko, and Patrick Ernst. 2024. Calibrating verbalized probabilities for large language models. Preprint, arXiv:2410.06707. — [arXiv:2410.06707](https://arxiv.org/abs/2410.06707) [find](https://scholar.google.com/scholar?q=Cheng+Wang%2C+Gyuri+Szarvas%2C+Georges+Balazs%2C+Pavel+Danchenko%2C+and+Patrick+Ernst%2E+2024%2E+Calibrating+verbalized+probabilities+for+large+language+models%2E+Preprint%2C+arXiv%3A2410%2E06707%2E+%E2%80%94+%5BarXiv%3A2410%2E06707%5D%28ht)
44. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, brian ichter, Fei Xia, Ed H. Chi, Quoc V Le, and Denny Zhou. 2022. [Chain of thought prompting elicits reasoning in large language models](https://md.research.elamur.ai/free/en/papers/wei-2022-chain-thought-prompting-elicits-reasoning/). In Advances in Neural Information Processing Systems.
45. Changyi Xiao, Mengdi Zhang, and Yixin Cao. 2025a. Bnpo: Beta normalization policy optimization. Preprint, arXiv:2506.02864. — [arXiv:2506.02864](https://arxiv.org/abs/2506.02864) [find](https://scholar.google.com/scholar?q=Changyi+Xiao%2C+Mengdi+Zhang%2C+and+Yixin+Cao%2E+2025a%2E+Bnpo%3A+Beta+normalization+policy+optimization%2E+Preprint%2C+arXiv%3A2506%2E02864%2E+%E2%80%94+%5BarXiv%3A2506%2E02864%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2506%2E02864%29)
46. Jiancong Xiao, Bojian Hou, Zhanliang Wang, Ruochen Jin, Qi Long, Weijie J Su, and Li Shen. 2025b. Restoring calibration for aligned large language models: A calibration-aware fine-tuning approach. In Forty-second International Conference on Machine Learning. [find](https://scholar.google.com/scholar?q=Jiancong+Xiao%2C+Bojian+Hou%2C+Zhanliang+Wang%2C+Ruochen+Jin%2C+Qi+Long%2C+Weijie+J+Su%2C+and+Li+Shen%2E+2025b%2E+Restoring+calibration+for+aligned+large+language+models%3A+A+calibration%2Daware+fine%2Dtuning+approach%2E+In+)
47. Tianyang Xu, Shujin Wu, Shizhe Diao, Xiaoze Liu, Xingyao Wang, Yangyi Chen, and Jing Gao. 2024. SaySelf: Teaching LLMs to express confidence with self-reflective rationales. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 5985–5998, Miami, Florida, US… [find](https://scholar.google.com/scholar?q=Tianyang+Xu%2C+Shujin+Wu%2C+Shizhe+Diao%2C+Xiaoze+Liu%2C+Xingyao+Wang%2C+Yangyi+Chen%2C+and+Jing+Gao%2E+2024%2E+SaySelf%3A+Teaching+LLMs+to+express+confidence+with+self%2Dreflective+rationales%2E+In+Proceedings+of+the+2024)

*Updated 2 November 2026*