Knowledge-driven automated prefabricated bridge modeling from natural language using LLM and RAG.

The drop in component parameter accuracy (54.2% points, from 100% to 45.8%) was substantially larger than that in cross-section parameter accuracy (37.5% points, from 100% to 62.5%), indicating that the RAG module plays an critical and highly significant role in accurately extracting construction constraint parameters such as diaphragm positioning and lifting reservation holes.

also in 132,142 other papers