Multiagent reinforcement learning for personalized product recommendation with large language model enhancement
Abstract
Personalized product recommendation is a central challenge in online platforms, where accurately capturing user preferences and adapting to dynamic behavior patterns directly impacts conversion rates and user satisfaction. Traditional collaborative filtering and content-based methods struggle with data sparsity, cold-start problems, and the inability to model complex sequential decision-making processes. This paper proposes MARL-Rec, a Multi-Agent Reinforcement Learning framework enhanced by Large Language Models for personalized recommendation. The framework employs multiple cooperative agents: a user intent analysis agent powered by a fine-tuned language model, a graph-based collaborative filtering agent that captures user-item interaction patterns, and a policy coordination agent that balances exploration and exploitation. We evaluate MARL-Rec on two real-world datasets, Amazon Product Reviews and Taobao User Behavior, and demonstrate that it achieves significant improvements over state-of-the-art baselines, with HR@10 gains of 7.6% and 9.6% respectively. Ablation studies confirm that both the multi-agent architecture and the LLM-based intent module contribute meaningfully to the overall performance.