AI research
The research paper "Dynamic Ontological Synthesis: Real--Time Knowledge Graph Construction for Logic--Grounded RAG" (arXiv:2606.22001), published on June 22, 2026, introduces a framework to overcome the static nature of traditional ontologies in RAG systems. Researchers at UC Berkeley and Microsoft Research propose "Dynamic Ontological Synthesis" (DOS), which enables Large Language Models to autonomously generate and refine formal logical structures during the retrieval phase. By utilizing a "Schema--as--a--Service" architecture, the system ensures that retrieved information is validated against a dynamically evolving logic base, significantly improving reasoning accuracy in rapidly changing data environments.