ChronoMedKG: A Temporally-Grounded Biomedical Knowledge Graph and Benchmark for Clinical Reasoning
Researchers have developed ChronoMedKG, a novel biomedical knowledge graph that incorporates temporal information crucial for clinical reasoning. Unlike existing static KGs, ChronoMedKG links disease associations to specific temporal components like onset windows or progression stages, drawing from over 460,000 evidence-linked triples. This temporal grounding significantly aids LLMs in answering complex clinical questions, rescuing a substantial portion of their failures on temporal reasoning tasks. AI
IMPACT Enhances LLM capabilities in complex clinical reasoning by providing temporal context, improving accuracy on long-tail failures.