Elamur Research⌕
☰
home / papers / 2026 / Robust Uncertainty Quantification for Self-Evolving Large Language Models via Continual Domain Pretraining

Robust Uncertainty Quantification for Self-Evolving Large Language Models via Continual Domain Pretraining

Paper · 2026 · Xiaofan Zhou and Lu Cheng

Summary

The paper tackles uncertainty quantification for LLMs that keep learning through continual domain pretraining (CDP), where a model is fine-tuned on one knowledge domain after another. Standard Conformal Prediction (CP) assumes calibration and test data are exchangeable. That assumption fails when test queries come from an unlabeled, unknown mix of domains, so coverage becomes uneven across domains. CP can also return very large, uninformative answer sets when the model meets questions it can no longer handle, for example after catastrophic forgetting.

The authors propose AR-NECP, which has two parts:

  • Non-exchangeable CP: transformer-based clustering estimates the domain proportions in the test set, and the buffer calibration data are then reweighted or resampled to match.
  • Adaptive rejection CP: a label-conditional mechanism, driven by normalized entropy, lets the model output "can't answer" when its competence shifts.

Experiments on TriviaQA, HotpotQA and MMLU (e.g., with Llama-3.1-8B-Instruct) show that coverage holds under domain shift and prediction sets get smaller. The gains are largest at high coverage levels. This matters for making self-evolving LLMs statistically trustworthy.

Key points

  • The authors describe this as the first CP study for LLMs under CDP. They provide theoretical analysis of coverage under domain distribution shift and of efficiency (Appendix A). Code is released at github.com/AlearZhou/CPCL.
  • TriviaQA is split into five domains (sc, ht, sp, en, gg). Quantiles of the uncertainty scores differ noticeably between these domains, which motivates the method.
  • Normalized entropy (NE) rejection gives smaller prediction sets than an MLP-confidence baseline. Examples:
    • whole distribution, α=0.05: 7.95 vs 8.66
    • en-dominated test set, α=0.10: 3.72 vs 4.51
  • The domain-count estimator is accurate: mean relative error stays between 0.01 and 0.11 across domains and test shifts.
  • Efficiency gains are small on MMLU, a 4-option task where oversized sets are not an issue. The authors list two limitations:
    • variance in the domain-count estimates
    • the need to resample calibration answers after every fine-tuning stage

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 Illinois ChicagoChicago · United States61 works · 20,107 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

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,320 works

Cite

Cite as Xiaofan Zhou and Lu Cheng. 2026. Robust Uncertainty Quantification for Self-Evolving Large Language Models via Continual Domain Pretraining. 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.388

@inproceedings{zhou2026robust,
  title = {{Robust Uncertainty Quantification for Self-Evolving Large Language Models via Continual Domain Pretraining}},
  author = {Xiaofan Zhou and Lu Cheng},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.388},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.388}
}

References

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

  1. 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.
  2. Aisha Alansari and Hamzah Luqman. 2025. Large language models hallucination: A comprehensive survey. arXiv preprint arXiv:2510.06265. — [arXiv:2510.06265](https://arxiv.org/abs/2510.06265) find
  3. Anastasios N Angelopoulos and Stephen Bates. 2021. A gentle introduction to conformal prediction and distribution-free uncertainty quantification. arXiv preprint arXiv:2107.07511. find
  4. Margarida Campos, António Farinhas, Chrysoula Zerva, Mário AT Figueiredo, and André FT Martins. 2024. Conformal prediction for natural language processing: A survey. Transactions of the Association for Computational Linguistics, 12:1497–1516. find
  5. Pierre Colombo, Telmo Pessoa Pires, Malik Boudiaf, Dominic Culver, Rui Melo, Caio Corro, Andre FT Martins, Fabrizio Esposito, Vera Lúcia Raposo, Sofia Morgado, and 1 others. 2024. Saullm-7b: A pioneering large language model for law. arXiv preprint arXiv:2403.03883. find
  6. Rui Gao and Weiwei Liu. Model uncertainty quantification by conformal prediction in continual learning. In Forty-second International Conference on Machine Learning. find
  7. Yu Gui, Ying Jin, and Zhimei Ren. 2024. Conformal alignment: Knowing when to trust foundation models with guarantees. Advances in Neural Information Processing Systems, 37:73884–73919. find
  8. Suchin Gururangan, Mike Lewis, Ari Holtzman, Noah A Smith, and Luke Zettlemoyer. 2021. Demix layers: Disentangling domains for modular language modeling. arXiv preprint arXiv:2108.05036. find
  9. Suchin Gururangan, Ana Marasović, Swabha Swayamdipta, Kyle Lo, Iz Beltagy, Doug Downey, and Noah A. Smith. 2020. Don’t stop pretraining: Adapt language models to domains and tasks. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 8342–8360, Online. A… find
  10. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021b. Measuring massive multitask language understanding. Proceedings of the International Conference on Learning Representations (ICLR).
  11. Roger A Horn and Charles R Johnson. 2012. Matrix analysis. Cambridge university press. find
  12. Fanding Huang, Guanbo Huang, Xiao Fan, Yi He, Xiao Liang, Xiao Chen, Qinting Jiang, Faisal Nadeem Khan, Jingyan Jiang, and Zhi Wang. 2026. Semantic-space exploration and exploitation in rlvr for llm reasoning. Preprint, arXiv:2509.23808. find
  13. Quzhe Huang, Mingxu Tao, Chen Zhang, Zhenwei An, Cong Jiang, Zhibin Chen, Zirui Wu, and Yansong Feng. 2023. Lawyer llama technical report. arXiv preprint arXiv:2305.15062.
  14. 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
  15. Mandar Joshi, Eunsol Choi, Daniel Weld, and Luke Zettlemoyer. 2017. triviaqa: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension. arXiv e-prints, arXiv:1705.03551.
  16. Sung Jae Jung, Hajung Kim, and Kyoung Sang Jang. 2024. Llm based biological named entity recognition from scientific literature. In 2024 IEEE International Conference on Big Data and Smart Computing (Big-Comp), pages 433–435. IEEE. find
  17. Bhawesh Kumar, Charlie Lu, Gauri Gupta, Anil Palepu, David Bellamy, Ramesh Raskar, and Andrew Beam. 2023. Conformal prediction with large language models for multi-choice question answering. arXiv preprint arXiv:2305.18404. — [arXiv:2305.18404](https://arxiv.org/abs/2305.18404) find
  18. Jiangtong Li, Yuxuan Bian, Guoxuan Wang, Yang Lei, Dawei Cheng, Zhijun Ding, and Changjun Jiang. 2023. Cfgpt: Chinese financial assistant with large language model. arXiv preprint arXiv:2309.10654. — [arXiv:2309.10654](https://arxiv.org/abs/2309.10654) find
  19. Rujun Han, Xiang Ren, and Nanyun Peng. 2020. Econet: Effective continual pretraining of language models for event temporal reasoning. arXiv preprint arXiv:2012.15283. find
  20. Shuo Li, Sangdon Park, Insup Lee, and Osbert Bastani. 2024. TRAQ: Trustworthy retrieval augmented question answering via conformal prediction. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1:… find
  21. Dan Hendrycks, Collin Burns, Steven Basart, Andrew Critch, Jerry Li, Dawn Song, and Jacob Steinhardt. 2021a. Aligning ai with shared human values. Proceedings of the International Conference on Learning Representations (ICLR).
  22. Kaizhao Liang. 2024. MMLU Auxiliary Trained Set Labelled by e5-mistral-7b-instruct. https://huggingface.co/datasets/kz919/mmlu-auxiliary-train-e5-mistral-7b-instruct. Accessed: Date of Access. ↗
  23. Tuve Löfström, Henrik Boström, Henrik Linusson, and Ulf Johansson. 2015. Bias reduction through conditional conformal prediction. Intelligent Data Analysis, 19(6):1355–1375. find
  24. Huanxin Sheng, Xinyi Liu, Hangfeng He, Jieyu Zhao, and Jian Kang. 2025. Analyzing uncertainty of llm-as-a-judge: Interval evaluations with conformal prediction. Preprint, arXiv:2509.18658. find
  25. Yuheng Lu, Bingshuo Qian, Caixia Yuan, Huixing Jiang, and Xiaojie Wang. 2025. Controlled low-rank adaptation with subspace regularization for continued training on large language models. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers… find
  26. Haizhou Shi, Zihao Xu, Hengyi Wang, Weiyi Qin, Wenyuan Wang, Yibin Wang, Zifeng Wang, Sayna Ebrahimi, and Hao Wang. 2024. Continual learning of large language models: A comprehensive survey. ACM Computing Surveys. find
  27. Shirong Ma, Shen Huang, Shulin Huang, Xiaobin Wang, Yangning Li, Hai-Tao Zheng, Pengjun Xie, Fei Huang, and Yong Jiang. 2023. Ecomgpt-ct: Continual pre-training of e-commerce large language models with semi-structured data. arXiv preprint arXiv:2312.15696. — [arXiv:2312.15696](https://arxiv.org/abs/2312.15696) find
  28. Christopher Mohri and Tatsunori Hashimoto. 2024. Language models with conformal factuality guarantees. arXiv preprint arXiv:2402.10978. — [arXiv:2402.10978](https://arxiv.org/abs/2402.10978) find
  29. Viktor Moskvoretskii, Maria Marina, Mikhail Salnikov, Nikolay Ivanov, Sergey Pletenev, Daria Galimzianova, Nikita Krayko, Vasily Konovalov, Irina Nikishina, and Alexander Panchenko. 2025. Adaptive retrieval without self-knowledge? bringing uncertainty back home. In Proceedings of the 63rd Annual Mee…
  30. Yujia Qin, Cheng Qian, Xu Han, Yankai Lin, Huadong Wang, Ruobing Xie, Zhiyuan Liu, Maosong Sun, and Jie Zhou. 2023. Recyclable tuning for continual pre-training. arXiv preprint arXiv:2305.08702. find
  31. Yujia Qin, Jiajie Zhang, Yankai Lin, Zhiyuan Liu, Peng Li, Maosong Sun, and Jie Zhou. 2022. Elle: Efficient lifelong pre-training for emerging data. arXiv preprint arXiv:2203.06311. find
  32. Victor Quach, Adam Fisch, Tal Schuster, Adam Yala, Jae Ho Sohn, Tommi S Jaakkola, and Regina Barzilay. 2023. Conformal language modeling. arXiv preprint arXiv:2306.10193. find
  33. Nitin Liladhar Rane, Abhijeet Tawde, Saurabh P Choudhary, and Jayesh Rane. 2023. Contribution and performance of chatgpt and other large language models (llm) for scientific and research advancements: a double-edged sword. International Research Journal of Modernization in Engineering Technology and… find
  34. Maxon Rubin-Toles, Maya Gambhir, Keshav Ramji, Aaron Roth, and Surbhi Goel. 2025. Conformal language model reasoning with coherent factuality. arXiv preprint arXiv:2505.17126. — [arXiv:2505.17126](https://arxiv.org/abs/2505.17126) find
  35. Glenn Shafer and Vladimir Vovk. 2008. A tutorial on conformal prediction. Journal of Machine Learning Research, 9(3). find
  36. Jiayuan Su, Jing Luo, Hongwei Wang, and Lu Cheng. 2024. API is enough: Conformal prediction for large language models without logit-access. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 979–995, Miami, Florida, USA. Association for Computational Linguistics. find
  37. Gemma Team, Thomas Mesnard, Cassidy Hardin, Robert Dadashi, Surya Bhupatiraju, Shreya Pathak, Laurent Sifre, Morgane Rivière, Mihir Sanjay Kale, Juliette Love, and 1 others. 2024. Gemma: Open models based on gemini research and technology. arXiv preprint arXiv:2403.08295. find
  38. Ryan J Tibshirani, Rina Foygel Barber, Emmanuel Candes, and Aaditya Ramdas. 2019. Conformal prediction under covariate shift. Advances in neural information processing systems, 32. find
  39. Mariya Toneva, Alessandro Sordoni, Remi Tachet des Combes, Adam Trischler, Yoshua Bengio, and Geoffrey J Gordon. 2018. An empirical study of example forgetting during deep neural network learning. arXiv preprint arXiv:1812.05159. — [arXiv:1812.05159](https://arxiv.org/abs/1812.05159) find
  40. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, and 1 others. 2023. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971.
  41. Jiaqi Wang, Enze Shi, Huawen Hu, Chong Ma, Yiheng Liu, Xuhui Wang, Yincheng Yao, Xuan Liu, Bao Ge, and Shu Zhang. 2025. Large language models for robotics: Opportunities, challenges, and perspectives. Journal of Automation and Intelligence, 4(1):52–64. find
  42. Chaoyi Wu, Weixiong Lin, Xiaoman Zhang, Ya Zhang, Weidi Xie, and Yanfeng Wang. 2024. Pmc-llama: toward building open-source language models for medicine. Journal of the American Medical Informatics Association, 31(9):1833–1843. find
  43. Xinbo Wu and Lav R Varshney. 2024. Transformer-based causal language models perform clustering. arXiv preprint arXiv:2402.12151. find
  44. Shijia Xu, Zhou Wu, Xiaolong Jia, Yu Wang, Kai Liu, and April Xiaowen Dong. 2026. Self-correcting rag: Enhancing faithfulness via mmkp context selection and nli-guided mcts. Preprint, arXiv:2604.10734.
  45. Yongyu Yan, Kui Xue, Xiaoming Shi, Qi Ye, Jingping Liu, and Tong Ruan. 2023. Af adapter: Continual pretraining for building chinese biomedical language model. In 2023 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pages 953– 957. IEEE. find
  46. Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William W. Cohen, Ruslan Salakhutdinov, and Christopher D. Manning. 2018. HotpotQA: A dataset for diverse, explainable multi-hop question answering. In Conference on Empirical Methods in Natural Language Processing (EMNLP).
  47. Fanghua Ye, Mingming Yang, Jianhui Pang, Longyue Wang, Derek F. Wong, Emine Yilmaz, Shuming Shi, and Zhaopeng Tu. 2024. Benchmarking LLMs via uncertainty quantification. In The Thirty-eight Conference on NIPS Datasets and Benchmarks Track. find
  48. Xiaofan Zhou, Baiting Chen, Yu Gui, and Lu Cheng. 2025. Conformal prediction: A data perspective. ACM Computing Surveys. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhou-2026-robust-uncertainty-quantification-self-evolving/