Modality-Guided Mixture of Graph Experts with Entropy-Triggered Routing for Multimodal Recommendation

Researchers introduce MAGNET, a new AI system for multimodal recommendation that combines user behavior, visual, and textual data through specialized graph neural network experts. The system uses entropy-triggered routing to automatically balance different data types and improve recommendations for sparse datasets and long-tail items.