Adaptive Hallucination Detection in Large Language Models Using Multi-Stage Confidence Estimation and Semantic Consistency Analysis

Authors

  • Maud Estey Department of Computer Science, Binghamton University, Binghamton, NY, USA. Author
  • Juhana Raty School of Information Technology, University of Cincinnati, Cincinnati, OH, USA. Author
  • Egbert Haering School of Computing, Clemson University, Clemson, SC, USA. Author

Keywords:

hallucination detection; semantic consistency; confidence calibration; selective prediction; retrieval-augmented generation; risk control

Abstract

Large language models can produce fluent responses that contain unsupported or contradictory claims, and single-signal detectors often mistake high token likelihood for factual reliability. This paper develops an adaptive hallucination detection framework that combines multi-stage confidence estimation with semantic consistency analysis. The framework treats hallucination as a latent relational property between a prompt, a generated response, and available evidence, rather than as a surface-level lexical defect. It integrates four complementary signals: sequence-level lexical confidence, meaning-cluster semantic entropy, sample-based semantic consistency, and claim-level evidence support. These signals are normalized into support scores and fused through reliability-weighted geometric pooling, so that unstable or poorly calibrated estimators receive lower influence. An ambiguity-sensitive threshold adjusts the acceptable risk according to retrieval volatility and semantic disagreement, and a three-way decision rule separates acceptance, flagging for verification, and rejection. The analytical contribution is a coherent decision architecture that links uncertainty estimation, semantic clustering, evidence alignment, and selective prediction within one risk-control logic. The discussion identifies boundary conditions that matter in practice: correlated samples can overstate consistency, semantic clustering can merge distinct claims, retrieval noise can mislead evidence scores, and calibration can drift under distribution shift. The framework provides a technically grounded basis for deploying hallucination detection as a calibrated, context-sensitive risk-control layer rather than as a static post hoc filter.

References

1. Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y. J., Madotto, A., & Fung, P. (2023). Survey of hallucination in natural language generation. ACM Computing Surveys, 55(12), 1–38. https://doi.org/10.1145/3571730

2. Maynez, J., Narayan, S., Bohnet, B., & McDonald, R. (2020). On faithfulness and factuality in abstractive summarization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (pp. 1906–1919). Association for Computational Linguistics. https://doi.org/10.18653/v1/2020.acl-main.173

3. Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W.-t., Rocktäschel, T., Riedel, S., & Kiela, D. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. In Advances in Neural Information Processing Systems (Vol. 33, pp. 9459–9474). Curran Associates.

4. Kadavath, S., Conerly, T., Askell, A., Henighan, T., Drain, D., Perez, E., Schiefer, N., Hatfield-Dodds, Z., DasSarma, N., Tran-Johnson, E., Johnston, S., El-Showk, S., Jones, A., Elhage, N., Hume, T., Chen, A., Bai, Y., Bowman, S., Fort, S., ... Kaplan, J. (2022). Language models (mostly) know what they know (arXiv:2207.05221). arXiv. https://arxiv.org/abs/2207.05221

5. Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. (2017). On calibration of modern neural networks. In Proceedings of the 34th International Conference on Machine Learning (Vol. 70, pp. 1321–1330). PMLR.

6. Wang, X., Wei, J., Schuurmans, D., Le, Q., Chi, E., Narang, S., Chowdhery, A., & Zhou, D. (2023). Self-consistency improves chain of thought reasoning in language models. In International Conference on Learning Representations.

7. Kuhn, L., Gal, Y., & Farquhar, S. (2023). Semantic uncertainty: Linguistic invariances for uncertainty estimation in natural language generation. In International Conference on Learning Representations.

8. Zhang, R. (2025, December). Uncertainty-Aware RAG: Trustworthy Generation and Risk Control. In 2025 5th International Symposium on Artificial Intelligence and Big Data (AIBDF) (pp. 526-531). IEEE.

9. Farquhar, S., Kossen, J., Kuhn, L., & Gal, Y. (2024). Detecting hallucinations in large language models using semantic entropy. Nature, 630(8017), 625–630. https://doi.org/10.1038/s41586-024-07421-0

10. Manakul, P., Liusie, A., & Gales, M. J. F. (2023). SelfCheckGPT: Zero-resource black-box hallucination detection for generative large language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (pp. 9004–9017). Association for Computational Linguistics.

11. Azaria, A., & Mitchell, T. (2023). The internal state of an LLM knows when it's lying. In Findings of the Association for Computational Linguistics: EMNLP 2023 (pp. 967–976). Association for Computational Linguistics.

12. Burns, C., Ye, H., Klein, D., & Steinhardt, J. (2023). Discovering latent knowledge in language models without supervision. In International Conference on Learning Representations.

13. Gal, Y., & Ghahramani, Z. (2016). Dropout as a Bayesian approximation: Representing model uncertainty in deep learning. In Proceedings of the 33rd International Conference on Machine Learning (Vol. 48, pp. 1050–1059). PMLR.

14. Lakshminarayanan, B., Pritzel, A., & Blundell, C. (2017). Simple and scalable predictive uncertainty estimation using deep ensembles. In Advances in Neural Information Processing Systems (Vol. 30, pp. 6402–6413). Curran Associates.

15. Ovadia, Y., Fertig, E., Ren, J., Nado, Z., Sculley, D., Nowozin, S., Dillon, J. V., Lakshminarayanan, B., & Snoek, J. (2019). Can you trust your model's uncertainty? Evaluating predictive uncertainty under dataset shift. In Advances in Neural Information Processing Systems (Vol. 32, pp. 13991–14002). Curran Associates.

16. Min, S., Krishna, K., Lyu, X., Lewis, M., Yih, W.-t., Koh, P. W., Iyyer, M., Zettlemoyer, L., & Hajishirzi, H. (2023). FActScore: Fine-grained atomic evaluation of factual precision in long form text generation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (pp. 12076–12100). Association for Computational Linguistics.

17. Es, S., James, J., Espinosa-Anke, L., & Schockaert, S. (2024). RAGAS: Automated evaluation of retrieval augmented generation. In Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations (pp. 150–158). Association for Computational Linguistics.

18. Lin, S., Hilton, J., & Evans, O. (2022). Teaching models to express their uncertainty in words. Transactions on Machine Learning Research. https://arxiv.org/abs/2205.14334

Downloads

Published

2026-10-28