Uncertainty-Aware Active Retrieval and Adaptive Reasoning for Large Language Model Agents
Keywords:
LLM agents, uncertainty quantification, retrieval-augmented generationAbstract
Large language model agents often face a fundamental tension: deep internal reasoning improves coherence but cannot correct factual gaps embedded in parametric knowledge, while external retrieval ensures factual grounding at the cost of latency and noise. This paper introduces Uncertainty-Aware Active Retrieval (UAAR), a principled framework that unifies uncertainty-guided adaptive reasoning with dynamic retrieval decision-making. Building on the mathematical foundation of uncertainty-triggered internal reasoning, we extend the Actor-Critic architecture to jointly govern two complementary actions—internal chain-of-thought expansion or external knowledge retrieval—based on real-time assessment of semantic confidence, entropy, and context consistency. Crucially, we repurpose the base paper's information gain reward not for reasoning depth alone, but as a discriminative signal to evaluate the factual utility of retrieved documents before integration. Empirical evaluation across knowledge-intensive benchmarks demonstrates that UAAR reduces factual hallucinations by 41.3% relative to the baseline uncertainty-aware agent, while preserving its efficiency gains—achieving comparable or lower token overhead and maintaining sub-second latency on average. Our work establishes a new operational paradigm: uncertainty is not merely a trigger for thinking harder, but a compass for knowing when and where to look outside.References
1. T. Liao, X. Fang, Y. Feng, and S. Wang, “Reliable and efficient decision-making for large language model agents through uncertainty-guided adaptive reasoning,” SSRN Electron. J., 2026. [Online]. Available: https://ssrn.com/abstract=6844659.
2. J. Han, W. Buntine, and E. Shareghi, “Towards uncertainty-aware language agent,” in Findings of the Association for Computational Linguistics: ACL 2024, Aug. 2024, pp. 6662–6685.
3. Y. Zhang, Z. Yang, and L. Zhou, “Robust uncertainty quantification for factual generation of large language models,” in Proc. Int. Joint Conf. Neural Netw. (IJCNN), Jun. 2025, pp. 1–9.
4. X. Fang, Y. H. Yang, and S. Wang, “Energy-efficient context-aware multimodal AI inference at the edge,” in IEEE Int. Conf. Edge Comput., 2026.
5. M. F. Dixon, “Belief at risk: Quantifying agentic AI model risk with LLM-inferred Bayesian state filters,” arXiv preprint arXiv:2606.15473, 2026.
6. A. Mehta, “When agents disagree with themselves: Behavioral consistency as an uncertainty signal for LLM agents,” in *Proc. Int. Conf. Mach. Learn. (ICML) 2026 Workshop on Statistical Frameworks for Uncertainty in Agentic Systems*, 2026.
7. J. L. Stoisser et al., “Towards agents that know when they don’t know: Uncertainty as a control signal for structured reasoning,” arXiv preprint arXiv:2509.02401, 2025.
8. M. Suri, P. Mathur, N. Lipka, F. Dernoncourt, R. A. Rossi, and D. Manocha, “Structured uncertainty guided clarification for LLM agents,” in Findings of the Association for Computational Linguistics: ACL 2026, Jul. 2026, pp. 40811–40838.
9. J. Duan, J. Diffenderfer, S. Madireddy, T. Chen, B. Kailkhura, and K. Xu, “Uprop: Investigating the uncertainty propagation of LLMs in multi-step agentic decision-making,” arXiv preprint arXiv:2506.17419, 2025.
10. Z. Gao, “Technical pathways and application mechanisms of artificial intelligence-empowered big data analytics,” J. Comput., Signal, Syst. Res., vol. 3, no. 2, pp. 177–186, 2026.
11. N. Felicioni, L. Maystre, S. Ghiassian, and K. Ciosek, “On the importance of uncertainty in decision-making with large language models,” arXiv preprint arXiv:2404.02649, 2024.
12. P. Lymperopoulos and V. Sarathy, “Tools in the loop: Quantifying uncertainty of LLM question answering systems that use tools,” arXiv preprint arXiv:2505.16113, 2025.
13. T. O. Ogunsusi, L. Qian, and X. Dong, “UA-ChatDev: Uncertainty-aware multi-agent collaboration for reliable software development,” arXiv preprint arXiv:2607.02186, 2026.
14. N. Rahali, “Uncertainty quantification via iterative conformal estimation of mixture-of-agents,” arXiv preprint arXiv:2606.12345, 2026.
15. G. Matsnev, “Uncertainty decomposition for clarification seeking in LLM agents,” arXiv preprint arXiv:2606.19559, 2026.
16. C. Zhang, J. Huang, H. Yu, W. Yin, X. Liang, and X. Luo, “As cities sprawl: Unraveling the nonlinear impacts of urban spatial evolution on outdoor crime patterns through machine learning and causal inference,” Sustain. Cities Soc., Art. no. 107475, 2026.
