Answers are grounded in this paper and cite their sources.
Under rigid regional water-resource constraints, water-saving governance requires not only efficiency evaluation but also conversion of diagnostic results into policy-grounded strategies. This study proposes the Multi-Agent System with Knowledge Graph Enhancement for Water-saving Strategy Generation (MAS-Graph), a framework that combines a multi-source heterogeneous water-saving knowledge graph with multi-agent collaboration. The graph integrates regional water-use indicators, policy texts, spatial jurisdiction information, and Four-Water and Four-Determination constraints, organizing them into numerical fact, policy semantic, and rule constraint layers. On this basis, slack-based measure data envelopment analysis (SBM-DEA) diagnoses water-use inefficiency, Policy Modeling Consistency (PMC) features guide policy basis retrieval, and five agents coordinate intent parsing, evaluation, retrieval, drafting, and compliance review. A large language model (LLM) supports intent understanding and text generation, while the graph-based review mechanism verifies factual consistency, spatial jurisdiction, and rigid constraints. Experiments using Henan Province water-use data and policy texts show that MAS-Graph improves strategy specificity, policy-basis adequacy, and compliance-review rigor over the base LLM and conventional retrieval-augmented generation methods, supporting smart water-saving governance.
No takes yet. Share an insight, caveat, or question.
Yang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: