Multi-agent Architectures Based on Large-scale Low-generation Language Models for Solving Complex Legal Problems: A Comparative Study
( Pp. 350-355)
More about authors
Roman V. Dushkin
senior lecturer at Department 22 “Cybernetics”; Nuclear Research University MEPhI
Scientific Research Nuclear University of MEPhI
Moscow, Russian Federation Vladimir N. Podoprigora Cand. Sci. (Econ.), head of the laboratory; Russian Economic University. G.V. Plekhanov
Plekhanov Russian University of Economics
Moscow, Russian Federation Alexey A. Kuzmin General Director
Ecosystem Digital Solutions LLC
Moscow, Russian Federation Kirill R. Dushkin analyst; A-Z Expert LLC
LLC "A-Ya expert"
Moscow, Russian Federation
Scientific Research Nuclear University of MEPhI
Moscow, Russian Federation Vladimir N. Podoprigora Cand. Sci. (Econ.), head of the laboratory; Russian Economic University. G.V. Plekhanov
Plekhanov Russian University of Economics
Moscow, Russian Federation Alexey A. Kuzmin General Director
Ecosystem Digital Solutions LLC
Moscow, Russian Federation Kirill R. Dushkin analyst; A-Z Expert LLC
LLC "A-Ya expert"
Moscow, Russian Federation
Abstract:
This article presents a comparative analysis of five multi-agent architectures based on large, low-generation language models for solving complex legal problems. The study was conducted on a specially prepared dataset of 25 questions of five difficulty levels on Russian family and civil law. Architectures of varying complexity were tested: from a simple lawyer-agent to extended ensembles with a dispatcher and a "jury" system. The main evaluation metrics were the average response quality score, token consumption, economic cost, and efficiency coefficient. The results revealed significant differences between the architectures: Option 5 demonstrated the best quality (6.44 points), but Option 1 proved the most effective with a coefficient of 49.46. Complex architectures required 10-15 times more tokens with an insignificant increase in quality. Analysis by complexity levels revealed that multi-agent systems are most effective for problematic situations and conflicts of laws, while simpler architectures are sufficient for typical tasks. The study provides scientifically based recommendations for selecting optimal architectural solutions for legal advisory systems, balancing quality and cost-effectiveness.
How to Cite:
Dushkin, R.V., Podoprigora, V.N., Kuzmin, A.A., & Dushkin, K.R. (2025). Multi-agent architectures based on large-scale low-generation language models for solving complex legal problems: A comparative study. Economic Problems and Legal Practice, 21(5), 350-355. DOI: 10.33693/2541-8025-2025-21-5-350-355. EDN: GGUZGY
Reference list:
Guo T. Large language model based multi-agents: A survey of progress and challenges / T. Guo, X. Chen, Y. Wang, R. Chang, S. Pei // arXiv preprint arXiv:2402.01680. URL: https://arxiv.org/abs/2402.01680 (date accessed: 23.06.2025).
Dushkin R.V., Andronov M.G. Hybrid design of artificial intelligent systems. Cybernetics and Programming. 2019. No. 4. Pp. 51–58. (In Rus.). DOI: 10.25136/2644-5522.2019.4.29809. EDN: OKAMBF.
Binyamin S.S. Multi-agent systems for resource allocation and scheduling in a smart grid / S. S. Binyamin, S. Ben Slama // Sensors. 2022. URL: https://www.mdpi.com/1424-8220/22/21/8099 (date accessed: 23.06.2025).
Dushkin R.V., Andronov M.G. Hybrid design of artificial intelligent systems. Cybernetics and Programming. 2019. No. 4. Pp. 51–58. (In Rus.). DOI: 10.25136/2644-5522.2019.4.29809. EDN: OKAMBF.
Binyamin S.S. Multi-agent systems for resource allocation and scheduling in a smart grid / S. S. Binyamin, S. Ben Slama // Sensors. 2022. URL: https://www.mdpi.com/1424-8220/22/21/8099 (date accessed: 23.06.2025).
Keywords:
multi-agent systems, large language models, legal problems, architectural solutions, token efficiency, economic optimization, response quality, junior BJMs, family law, civil law..
