TY - GEN

T1 - MARL-GPT: Foundation Model for Multi-Agent Reinforcement Learning

AU - Иванова, Ольга Ярославна

AU - Яковлев, Константин Сергеевич

PY - 2026/5/24

Y1 - 2026/5/24

N2 - Recent advances in multi-agent reinforcement learning (MARL)have demonstrated success in numerous challenging domains and environments, but typically require specialized models for each task. In this work, we propose a coherent methodology that makes it possible for a single GPT-based model to learn and perform well across diverse MARL environments and tasks, including StarCraft Multi-Agent Challenge, Google Research Football and POGEMA. Our method, MARL-GPT, applies offline reinforcement learning to train at scale on the expert trajectories (400M for SMACv2, 100M for GRF, and 1B for POGEMA) combined with a single transformerbased observation encoder that requires no task-specific tuning.Experiments show that MARL-GPT1 achieves competitive performance compared to specialized baselines in all tested environments. Thus, our findings suggest that it is, indeed, possible to build a multitask transformer-based model for a wide variety of (significantly different) multi-agent problems paving the way to the fundamental MARL model (akin to ChatGPT, Llama, Mistral etc. in natural language modeling).

AB - Recent advances in multi-agent reinforcement learning (MARL)have demonstrated success in numerous challenging domains and environments, but typically require specialized models for each task. In this work, we propose a coherent methodology that makes it possible for a single GPT-based model to learn and perform well across diverse MARL environments and tasks, including StarCraft Multi-Agent Challenge, Google Research Football and POGEMA. Our method, MARL-GPT, applies offline reinforcement learning to train at scale on the expert trajectories (400M for SMACv2, 100M for GRF, and 1B for POGEMA) combined with a single transformerbased observation encoder that requires no task-specific tuning.Experiments show that MARL-GPT1 achieves competitive performance compared to specialized baselines in all tested environments. Thus, our findings suggest that it is, indeed, possible to build a multitask transformer-based model for a wide variety of (significantly different) multi-agent problems paving the way to the fundamental MARL model (akin to ChatGPT, Llama, Mistral etc. in natural language modeling).

KW - Multi-Task Learning

KW - Multi-agent Learning

KW - Reinforcement Learning

UR - https://www.mendeley.com/catalogue/14016dab-adae-35e2-bd7c-69d0a3240ccd/

U2 - 10.65109/bwfp6427

DO - 10.65109/bwfp6427

M3 - статья в сборнике материалов конференции

SN - 9798400723179

SP - 3843

EP - 3851

BT - AAMAS 2026: Proc. of the 25th International Conference on Autonomous Agents and Multiagent Systems

PB - Association for Computing Machinery

Y2 - 25 May 2026 through 29 May 2026

ER -