A METHOD FOR CONTROLLED ITERATIVE PROCESS EXECUTION IN SYSTEMS BASED ON LARGE LANGUAGE MODELS
DOI:
https://doi.org/10.28925/2663-4023.2026.34.1308Keywords:
large language models, interactive models, agent-based systems, prompt engineering, process control, semantic networks, process convergence, adaptive stopping conditionAbstract
This paper proposes a method for the controlled iterative execution of processes in systems based on large language models (LLMs), which is based on the composition of the ‘Function’, ‘Supercycle’ and ‘Condition’ primitives. It is shown that a single execution of queries to an LLM in tasks of knowledge extraction, text analysis and the construction of structured data representations is fundamentally incomplete due to the stochastic nature of result generation, the limitations of the context window and the dependence of the result on the preceding context. The proposed approach implements a controlled iterative mechanism for accumulating results with novelty control and an adaptive termination condition. A mathematical model of the dynamics of information gain has been developed, according to which the number of new results at each iteration exhibits a steady downward trend and can be approximated by a Gaussian function. On this basis, the convergent nature of the process is substantiated and the concept of ε-completeness of a result is introduced, under which further iterations do not provide significant information gain. Experimental verification has confirmed the stability of the proposed mechanism and the effectiveness of the adaptive termination condition. The results obtained provide a methodological basis for constructing controllable LLM-oriented systems capable of implementing multi-stage processes of analysis, knowledge extraction, semantic structure formation, and support for agent-based scenarios with controlled completeness and predictable termination
Downloads
References
Liu, P., Yuan, W., Fu, J., Jiang, Z., Hayashi, H., & Neubig, G. (2023). Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing. ACM Computing Surveys, 55(9), 1-35. https://doi.org/10.1145/3560815
Aghajanyan, A., Okhonko, D., Lewis, M., Joshi, M., Xu, H., Ghosh, G., & Zettlemoyer, L. (2021). HTLM: Hyper-text pre-training and prompting of language models [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2107.06955
Sahoo, P., Singh, A. K., Saha, S., Jain, V., Mondal, S., & Chadha, A. (2024). A systematic survey of prompt engineering in large language models: Techniques and applications [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2402.07927
Lande, D., Yefremov, K., Soboliev, A., & Pyshnograiev, I. (2025). Devising a code-free method for detecting signs of informational-psychological influences in messages. Eastern-European Journal of Enterprise Technologies, 5(2(137)), 55-69. https://doi.org/10.15587/1729-4061.2025.342297
Tang, J., Fan, T., & Huang, C. (2025). AutoAgent: A fully automated and zero-code framework for LLM agents [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2502.05957
Irmak, G., & Mahmoud, Q. H. (2026). Design behaviour and interface consistency in generative no-code tools: A systematic literature review. Computers, 15(4), Article 238. https://doi.org/10.3390/computers15040238
Lande, D., Danyk, Y., & Strashnoy, L. (2026). A no-code programming framework with self-correction based on large language models. In P. Venherskyi, S. Yevseiev, O. Gutik, V. Trushevskyy, & M. Viachalo (Eds.), Proceedings of the Workshop on Cryptology and Data Security (WCDS 2025), co-located with SMICS 2025 (CEUR Workshop Proceedings, Vol. 4191, pp. 22-37). CEUR-WS.org. https://ceur-ws.org/Vol-4191/paper1.pdf
Lande, D., & Strashnoy, L. (2025a). An advanced no-code programming framework for complex problems in LLM environments [Preprint]. ResearchGate. https://doi.org/10.13140/RG.2.2.25307.89129
Lande, D., & Strashnoy, L. (2025b). Orchestration of no-code agents in the AgentFlow framework [Preprint]. SSRN. https://doi.org/10.2139/ssrn.5381820
Ланде, Д., & Рибак, О. (2025). Безкодовий підхід до побудови семантичних мереж засобами промпт-інжинірингу. Information Technology and Security, 13(2), 264-278. https://doi.org/10.20535/2411-1031.2025.13.2.344712
Xu, W., Zhu, G., Zhao, X., Pan, L., Li, L., & Wang, W. (2024). Pride and prejudice: LLM amplifies self-bias in self-refinement. In L.-W. Ku, A. Martins, & V. Srikumar (Eds.), Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics: Volume 1. Long Papers (pp. 15474-15492). Association for Computational Linguistics. https://doi.org/10.18653/v1/2024.acl-long.826
Madaan, A., et al. (2023). Self-refine: Iterative refinement with self-feedback. Advances in Neural Information Processing Systems, 36, 46534-46594. https://proceedings.neurips.cc/paper_files/paper/2023/file/91edff07232fb1b55a505a9e9f6c0ff3-Paper-Conference.pdf
Shinn, N., Cassano, F., Gopinath, A., Narasimhan, K., & Yao, S. (2023). Reflexion: Language agents with verbal reinforcement learning. Advances in Neural Information Processing Systems, 36, 8634-8652. https://proceedings.neurips.cc/paper_files/paper/2023/file/1b44b878bb782e6954cd888628510e90-Paper-Conference.pdf
Kumar, A., et al. (2024). SCoRe: Self-correcting reasoning in large language models [Preprint]. arXiv. https://arxiv.org/pdf/2511.09381
Zelikman, E., Lorch, E., Mackey, L., & Kalai, A. T. (2024). Self-taught optimizer (STOP): Recursively self-improving code generation. In First Conference on Language Modeling. OpenReview. https://openreview.net/forum?id=46Zgqo4QIU
Rafique, Z., Wasim, M., Hussain, M., Hussain, M., & Memon, M. I. (2025). From prompt to action: A comprehensive review of LLM autonomous agents. In 2025 IEEE International Conference on Wireless for Space and Extreme Environments (WiSEE) (pp. 1-6). IEEE. https://doi.org/10.1109/WiSEE57913.2025.11229863
Fan, W., Ding, Y., Ning, L., Wang, S., Li, H., Yin, D., Chua, T.-S., & Li, Q. (2024). A survey on RAG meeting LLMs: Towards retrieval-augmented large language models. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (pp. 6491-6501). Association for Computing Machinery. https://doi.org/10.1145/3637528.3671470
Lande, D., Puchkov, O., & Subach, I. (2021). Aggregation of information from diverse networks as the basis for training cyber security specialists on processing ultra large data sets. Information Technology and Security, 9(1), 4-16. https://doi.org/10.20535/2411-1031.2021.9.1.247256
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Олександр Пучков, Дмитро Ланде, Ігор Субач
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
