Abstract: An explanatory rule mining framework for knowledge graphs based on large language models, Gear-LLMs, is proposed and its military applications are explored. The framework, consisting of such three steps as rule sampling, model rule generation, and rule verification, the natural language processing capability of large language models is utilized to effectively produce a great number of logical rules. The experimental results show that Gear-LLMs achieves the optimal results on the large-scale knowledge graph YAGO and significantly enhances the effects of knowledge graph completion tasks, verifies the significant role of large language models in logical rule mining.