LogosKG: Scalable and Interpretable Knowledge Graph Retrieval
LogosKG: Hardware-Optimized Knowledge Graph Retrieval
LogosKG provides efficient, scalable, and interpretable multi-hop retrieval over large knowledge graphs. It uses matrix-based representations and GPU acceleration to support biomedical knowledge graphs at billion-edge scale.
The project integrates knowledge graph retrieval with large language models to support evidence-grounded responses and improve reliability in biomedical applications such as clinical diagnosis verification and decision support.
Publication
Published at the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026).