LLM framework improves BIM design defect identification to 85% accuracy

A new framework uses BIM-to-Text chunking and RAG with rule injection to identify design defects in Building Information Modeling. The approach achieves 85% identification accuracy and 94% reasonable repair suggestions while reducing hallucinations by 92.5%.