Yiling Yuan
Information Networking Institute, Carnegie Mellon University, Pittsburgh, PA 15213, USA.
*Corresponding author: Yiling Yuan
Abstract
Large language models are very important in constructing intelligent agents with excellent natural language processing and logical reasoning capabilities, and the enhancement of the intelligent autonomous decision-making ability depends on the optimization of the agent’s memory management and dynamic reasoning mechanisms. In this research, we thoroughly study the basic technical principles of large language models based on the transformer architecture and pre-training alignment methods. A complete agent’s memory management system, comprising hierarchical memory modelling, multi-level information storage and retrieval, and adaptive memory update and forgetting optimization, has been established. Additionally, a dynamic reasoning system suitable for complex situations has been developed, which can accomplish scenario-adaptive matching, iterative reasoning based on memory, reduction of logical errors, and adjustment of environmental strategies by itself. The designed technical system can overcome the shortcomings of unsatisfactory memory management and poor dynamic reasoning adaptability in traditional intelligent agents, and significantly improve the environmental adaptability, reasoning accuracy, and task execution stability of large language model intelligent agents in different kinds of complicated applications.
References
[1] Li T, Ge X, Wang Z, et al. Multimodal large language model (MLLM) benchmark for intelligent construction in underground engineering. Autom Constr. 2026;188:106997.
doi:10.1016/j.autcon.2026.106997
[2] Zhu Z, Zhang J, Zhao S, et al. Large language model-enhanced causal Bayesian network: an intelligent modeling approach to maritime accident risk analysis. Ocean Eng. 2026;359(P3):126059. doi:10.1016/j.oceaneng.2026.126059
[3] Pape JA, Kulkarni S, Fregoli D, et al. Current artificial intelligence large language models exhibit sycophantic behavior in ortho-paedic contexts. J Bone Joint Surg. 2026.
doi:10.2106/JBJS.25.01576
[4] Hlaing MMHN, Park K, Wan Q, et al. Accuracy and hallucination behaviors of artificial intelligence-based large language models for dental implant identification on radiographs. J Dent. 2026;106754. doi:10.1016/j.jdent.2026.106754
[5] Poudel A, Barrios C, Torre LDP, et al. Governance risks of AI reasoning in urban infrastructure through Delphi audit of human and large language model judgment. Discov Cities. 2026;3(1):81.
doi:10.1007/s44327-026-00268-2
[6] Chuan HC, Tang J, Yang Z, et al. Evaluating AI-mediated health communication via large language model-based frequently asked questions rewriting to foster clinical trial participation: comparative survey study. JMIR AI. 2026;5:e87446.
doi:10.2196/87446
[7] Tang Z, Wu S, Zhao X, et al. A chromatographic process design and optimization multi-agent platform powered by large language models: a case application on extract of Ginkgo biloba leaf. J Chromatogr A. 2026;1781:467096.
doi:10.1016/j.chroma.2026.467096
[8] Erdogan H, Cankar M. Letter to the Editor: comment on “Evaluating the evidence-based potential of six large language models in paediatric dentistry: a comparative study on generative artificial intelligence”. Eur Arch Paediatr Dent. 2026.
doi:10.1007/s40368-026-01220-z
How to cite this paper
Research on Memory Management and Dynamic Reasoning Methods for Intelligent Agents Based on Large Language Models
How to cite this paper: Yiling Yuan. (2026). Research on Memory Management and Dynamic Reasoning Methods for Intelligent Agents Based on Large Language Models. Engineering Advances, 6(2), 135-138.
DOI: http://dx.doi.org/10.26855/ea.2026.06.011