Odin: Multi-Signal Graph Intelligence for Autonomous Discovery in Knowledge Graphs

Researchers present Odin, the first production-deployed graph intelligence engine that autonomously discovers patterns in knowledge graphs without predefined queries. The system uses a novel COMPASS scoring metric combining structural, semantic, temporal, and community-aware signals, and has been successfully deployed in regulated healthcare and insurance environments.

• →Odin is the first autonomous graph intelligence system deployed in production for regulated industries like healthcare and insurance.
• →The COMPASS scoring system combines four signals: structural importance, semantic plausibility, temporal relevance, and community-aware guidance.
• →The system addresses the 'echo chamber' problem where graph exploration gets trapped in dense local communities through bridge scoring mechanisms.
• →Beam search with multi-signal guidance achieves O(b·h) complexity while maintaining high recall compared to exhaustive exploration.
• →Complete provenance traceability is maintained, which is critical for regulated industries where hallucination is unacceptable.