Efficient and Scalable Retrieval in Knowledge Graphs: From Single-Graph to Partitioned Multi-Graph Approaches

This lightning talk introduces LogosKG, a framework for efficient and scalable multi-hop knowledge graph retrieval. The approach leverages sparse incidence factorizations for single-graph retrieval and extends to partitioned multi-graph designs with degree-aware routing, caching, and distributed execution. Applications in large-scale biomedical reasoning and LLM integration are discussed.