LLM-Augmented Multi-Agent Reinforcement Learning for Cross-Scenario Knowledge Transfer
Entropy · 06.05.2026 · 28 (5) · 525
DOI: 10.3390/e28050525 ↗Kurz gefasst
Fachartikel in der Zeitschrift Entropy, online erschienen am 06.05.2026. Verfasst von Chao Li und 5 weiteren Autor:innen. Beteiligte Einrichtung: Department of Basic Courses, Rocket Force University of Engineering, Xi'an 710025, China. Die Arbeit nennt 45 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Natural Science Basic Research Program of Shaanxi Province.
Abstract (Originalsprache)
Multi-agent reinforcement learning (MARL) relies on trial-and-error interactions to update policies. However, trial-and-error learning typically requires extensive interactions to achieve satisfactory performance, resulting in low sample efficiency, which limits its application in the real world. To reduce the trial-and-error costs of MARL and accelerate the convergence of multi-agent collaborative policies, we propose a MARL policy transfer method named LoLM-MARL, based on fine-tuning large language models (LLMs). First, leveraging the general world knowledge and reasoning capabilities of LLMs, low-rank adaptation (LoRA) is employed to fine-tune the pre-trained model on source tasks, thereby providing general decision-making knowledge for cross-scenario policy transfer. Second, a dynamic prompt construction method for LLMs is designed. By dynamically eliminating the state information of ineffective agents from the prompts, the method provides denser observation data for the large language model, thereby enhancing its policy representation capability in specific complex collaborative scenarios. Meanwhile, the dynamic prompt design concept enriches the training sub-scenarios for the algorithm, thereby laying the foundation for the model to learn more general decision-making knowledge. Finally, a Kullback–Leibler (KL) divergence regularization method based on an annealing strategy is constructed to ensure consistency between the policy distributions of the fine-tuned model and the pre-trained model, effectively mitigating the catastrophic forgetting problem during the fine-tuning process of the pre-trained model. Experimental results show that in zero-shot transfer tasks, LoLM-MARL achieves a maximum improvement of 101.4% in average win rate compared to existing state-of-the-art (SOTA) methods. In six few-shot transfer tasks, our method consistently achieves better generalization performance than traditional SOTA methods, and improves the convergence speed by 4 to 30 times compared to the training-from-scratch approach, providing a new solution paradigm for efficient policy transfer in complex dynamic environments.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Basic Courses, Rocket Force University of Engineering, Xi'an 710025, China.
- Department of Optical Communication Networks, Information Support Force Engineering University, Wuhan 430035, China.
ORCID-Profile: Chao Li
Förderung
- National Natural Science Foundation of China
- Natural Science Basic Research Program of Shaanxi Province
Zitieren
APA
Li, C., Liu, Y., Wang, J., Wang, Z., Lu, K., & Wang, C. (2026). LLM-Augmented Multi-Agent Reinforcement Learning for Cross-Scenario Knowledge Transfer. Entropy, 28(5), 525. https://doi.org/10.3390/e28050525
Vancouver
Li C, Liu Y, Wang J, Wang Z, Lu K, Wang C. LLM-Augmented Multi-Agent Reinforcement Learning for Cross-Scenario Knowledge Transfer. Entropy. 2026;28(5):525. doi:10.3390/e28050525
BibTeX
@article{Li2026,
author = {Li, Chao and Liu, Yanfei and Wang, Jieling and Wang, Zhong and Lu, Kewei and Wang, Chengjin},
title = {LLM-Augmented Multi-Agent Reinforcement Learning for Cross-Scenario Knowledge Transfer},
journal = {Entropy},
year = {2026},
volume = {28},
number = {5},
pages = {525},
doi = {10.3390/e28050525}
}
EUVIMED