Evaluating Fine-Tuned LLMs with RAG and Text2SQL for the Water Conservancy Domain

This study presents WaterQA, an intelligent question-answering platform addressing challenges in water resource management using Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG). By integrating RAG for accurate information retrieval from a dynamic knowledge hub and employing fine-tuned Text2SQL for querying real-time water level databases, the system enhances decision support. Supervised fine-tuning of the model demonstrated superior performance on complex, domain-specific tasks compared to other models. WaterQA, also incorporating water level prediction, offers a practical and accurate solution, advancing smart water conservancy through enhanced data accessibility and analysis.