Guiding Federated Graph Recommendation with LLM-encoded knowledge
Researchers have developed a new framework that leverages Large Language Models (LLMs) to enhance federated graph recommendation systems. This approach addresses the challenge of aggregating structural embeddings across distributed, non-IID clients in federated learning by using LLM-encoded knowledge to guide the process. Clients learn local graph representations and summarize interaction patterns into semantic vectors via a frozen LLM, which the central server then uses to discover related preference patterns and selectively aggregate structural representations. Experiments show this method improves recommendation accuracy over existing federated graph baselines. AI
IMPACT Enhances privacy-preserving recommendation systems by integrating LLM semantic understanding with federated learning.