Calibrated Abstention and Escalation in Foundation-Model Systems

Authors

  • Anthony Young Author

Keywords:

large language models, abstention, calibration, uncertainty, human escalation

Abstract

This review examines calibrated abstention and escalation in foundation-model systems. The organizing question is when a model should answer, defer, request evidence, or transfer control to a person. Ten related scholarly sources are synthesized through a decision-centered framework spanning problem definition, mechanism, measurement, evaluation, implementation, and governance. The review does not invent experiments, pooled estimates, or unreported quantitative results. It instead evaluates the strength and transferability of the available evidence, with particular attention to treating a fluent refusal style as evidence of calibrated uncertainty. The resulting framework links technical or empirical performance to explicit use conditions and identifies tests that should precede wider adoption in high-consequence decision support with bounded model authority.

References

  1. Ehrlich-Sommer, F., Eberhard, B., & Holzinger, A. (2025). ForestGPT and Beyond: A Trustworthy Domain-Specific Large Language Model Paving the Way to Forestry 5.0. Electronics, 14(18), 3583. https://doi.org/10.3390/electronics14183583
  2. Geng, J., Cai, F., Wang, Y., Koeppl, H., Nakov, P., & Gurevych, I. (2023). A Survey of Confidence Estimation and Calibration in Large Language Models. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2311.08298
  3. Hu, M., Zhang, Z., Zhao, S., Huang, M., & Wu, B. (2023). Uncertainty in Natural Language Processing: Sources, Quantification, and Applications. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2306.04459
  4. Lin, Z., Trivedi, S., & Sun, J. (2023). Generating with Confidence: Uncertainty Quantification for Black-box Large Language Models. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2305.19187
  5. Lu, Y., Chen, T., Hao, N., Rechem, C. V., Chen, J., & Fu, T. (2024). Uncertainty Quantification and Interpretability for Clinical Trial Approval Prediction. Health Data Science, 4, 0126. https://doi.org/10.34133/hds.0126
  6. Mena, J., Pujol, O., & Vitria, J. (2020). Uncertainty-Based Rejection Wrappers for Black-Box Classifiers. IEEE Access, 8, 101721-101746. https://doi.org/10.1109/access.2020.2996495
  7. Ren, J., Zhao, Y., Vu, T., Liu, P. J., & Lakshminarayanan, B. (2023). Self-Evaluation Improves Selective Generation in Large Language Models. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2312.09300
  8. Shorinwa, O., Mei, Z., Lidard, J., Ren, A. Z., & Majumdar, A. (2025). A Survey on Uncertainty Quantification of Large Language Models: Taxonomy, Open Research Challenges, and Future Directions. ACM Computing Surveys, 58(3), 1-38. https://doi.org/10.1145/3744238
  9. Tomani, C., Chaudhuri, K., Evtimov, I., Cremers, D., & Ibrahim, M. (2024). Uncertainty-Based Abstention in LLMs Improves Safety and Reduces Hallucinations. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2404.10960
  10. Wen, B., Yao, J., Feng, S., Xu, C., Tsvetkov, Y., Howe, B., & Wang, L. L. (2025). Know Your Limits: A Survey of Abstention in Large Language Models. Transactions of the Association for Computational Linguistics, 13, 529-556. https://doi.org/10.1162/tacl_a_00754

Downloads

Published

2026-08-05

Issue

Vol. 1 (2026): Archive

Section

Review

License

Copyright (c) 2026 Machine Intelligence & Responsible Systems

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.