[Purpose/significance] In the context of data-driven governance, solving the problem of fragmented govern⁃ment data governance policies and achieving overall policy effectiveness improvement have become key breakthroughs in unleashing the value of data elements and building a modern governance system. [Method/process] This article is based on the third-order research paradigm of "dynamic evolution analysis→ collaborative linkage mining → efficien⁃cy improvement path extraction". On the basis of constructing the theoretical framework of government data governance policy efficiency improvement，relying on the GraphRAG framework and integrating the deep semantic reasoning abili⁃ty of the big language model with the multidimensional relationship mapping technology of the knowledge graph, this ar⁃ticle analyzes the evolution process and evolution law of China's government data governance policy. Furthermore, it an⁃alyzes the collaboration of policy subjects and the linkage of policy content elements, explores the collaborative difficul⁃ties of China's government data governance policy, and finally proposes efficiency improvement path suggestions, pro⁃viding a practical direction of "evolution interpretation collaborative optimization efficiency improvement" for optimiz⁃ing government data governance policy. [Result/conclusion] China's government data governance policies have evolved from "block segmentation" to "multi-dimensional collaborative ecology", from "single point breakthrough" to "system empowerment", and from "passive response" to "value consciousness" in cognition; there are collaborative challenges such as "cellular isolation" and "breadth intensity imbalance"; it is necessary to build a cross domain collab⁃orative governance ecosystem, improve the dynamic adaptation mechanism of policy content, and enhance the overall synergy of multidimensional policies in order to improve the overall effectiveness of policies.