LLM-guided federated graph recommendation boosts accuracy across non-IID clients

Graph-based recommender systems are highly effective at extracting collaborative signals from user-item interactions, and federated learning (FL) allows these models to be trained while preserving user privacy. However, aggregating graph representations across distributed, non-IID clients remains a challenge. Structural embeddings learned locally often misalign, and naive averaging fails to capture meaningful cross-client relationships. Most existing federated graph methods rely exclusively on structural aggregation, neglecting the rich, global semantic context available in large language models (LLMs).

To address this, the authors propose using LLM-encoded knowledge to guide federated graph recommendation. Specifically, clients learn structural representations from local graphs while simultaneously summarizing their typical interaction patterns into compact semantic vectors via a frozen LLM. The central server then uses these semantic signals to discover related preference patterns across clients, guiding the selective aggregation of structural representations. This enables semantically informed cross-client collaboration without exposing raw data. The LLM is frozen, avoiding additional training overhead.

Extensive experiments on standard benchmarks show that guiding structural alignment with LLM-encoded knowledge consistently improves recommendation accuracy over existing federated graph baselines. This work bridges graph-based recommendation and LLM semantic understanding, offering a practical, privacy-preserving approach for real-world federated systems where user data is siloed and distributions differ across clients.