Implications of Using Large Language Models and Knowledge Graphs for Query Accuracy

Leveraging knowledge graphs makes large language models (LLMs) create more accurate queries for both easy and hard questions, along with easy and complex schemas. This is especially impactful for the most complex queries, questions, and schemas. The approach taken in this research is open and transparent, allowing others to extend and leverage the benchmark to explore alternative methods such as advanced ERD diagrams or encoding knowledge in plain English. The goal is to establish a community resource for empirical answers to questions regarding the effectiveness of different methods.