Traffic Safety RAG Paper

A research paper published on arXiv proposing a Retrieval-Augmented Generation framework that automatically converts accident narrative text into intersection safety countermeasure recommendations. The approach combines embedding-based retrieval, association rule mining, and engineering reasoning guidance, achieving an F1 score of 0.82 on real-world data.