LLM-BASED AGENT ARCHITECTURE: A NEW PARADIGM OF ARTIFICIAL INTELLIGENCE IN SOFTWARE ENGINEERING AUTOMATION
Received: 2026-05-31
Published: 2026-06-06
Abstract
This paper presents a scientific and analytical investigation into the architecture of Large Language Model (LLM)-based agents and their transformative role in software engineering. The ReAct, Chain-of-Thought, and Reflexion paradigms are comparatively evaluated for their effectiveness in resolving complex engineering tasks. The architectural principles of multi-agent systems and tool-augmented agents are analyzed alongside their integration methods across the Software Development Life Cycle (SDLC). Drawing on empirical benchmarks from platforms including GitHub Copilot, Amazon CodeWhisperer, Devin, and OpenClaw, the study develops practical recommendations for Uzbekistan’s software industry. Findings indicate that properly architected LLM agents can increase developer productivity by 35–65% and substantially reduce bug-resolution cycles.
Keywords
List of references
-
Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., & Wen, J. (2024). A Survey on Large Language Model-based Autonomous Agents. Frontiers of Computer Science. Vol.†18. No.†66. P. 186345. https://doi.org/10.1007/s11704-024-40231-1
-
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Chi, E., Le, Q., & Zhou, D. (2022). Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. Advances in Neural Information Processing Systems (NeurIPS). Vol.†35. P. 24824–24837.
-
Ernst, N. A., Carver, J., & Meng, N. (2022). Technical Debt and Refactoring: A Large-Scale Empirical Study. IEEE Transactions on Software Engineering. Vol.†49. No.†3. P. 1201–1219. https://doi.org/10.1109/TSE.2022.3152459
-
Liu, Y., He, H., Han, T., Zhang, X., Liu, M., & Ge, Z. (2024). Understanding LLM Hallucinations in Code Generation. IEEE Software. Vol.†41. No. 2. P. 88–97. https://doi.org/10.1109/MS.2023.3336800
-
Russell, S., & Norvig, P. (2020). Artificial Intelligence: A Modern Approach (4th ed.). Pearson. Upper Saddle River, NJ. 1132 p.
About the Authors
License
Copyright (c) 2026 MMIT Proceedings
This work is licensed under a Creative Commons Attribution 4.0 International License.