| 138 | 0 | 12 |
| Downloads | Citas | Reads |
With the continuous development of large-scale model technology, its application value in the military field is increasingly prominent. To explore the technical architecture and development process suitable for military applications of large-scale models and further enhance combat effectiveness, based on an in-depth analysis of the military applications requirements for large-scale models, a systematic architecture development process of "pretraining—supervised fine-tuning—reward modeling—reinforcement learning— knowledge distillation" is proposed by studying the technical mechanisms and military applicability of Transformer and Mixture of Experts(MoE) architectures. The feasibility of the development process is verified by taking the intelligent situation awareness system for joint maritime operations as an example. The results clearly demonstrate the potential application value of large-scale models in tasks such as multi-source heterogeneous data fusion and intelligent decision support, indicating the feasibility of the development process. The research provides theoretical references for promoting the in-depth application of large-scale models in the military field and accelerating the intelligent transformation of the military.
[1]郭旺，杨雨森，吴华瑞，等.农业大模型：关键技术、应用分析与发展方向[J].智慧农业（中英文），2024,6(2):1-13.
[2]崔翛龙，高志强，姬纬通，等.“艾武大模型+”：一种大模型军事化应用系统的开发与实证[J].数据采集与处理，2024,39(3):588-597.
[3]蔡磊，孟宪波，韩冬梅，等.大模型在军事垂直领域的应用[J].舰船科学技术，2024,46(5):171-175.
[4]徐皮克.面向军事知识问答的问题语义解析关键技术研究[D].长沙：国防科技大学，2021.
[5]Pfaff C A,Lowrance C J,Washburn B M,et al. Trusting AI:Integrating artificial intelligence into the army’s professional expert knowledge[J]. Journal of Defense Analytics,2023,15(3):45-60.
[6]刘涛，蒋国权，丁鲲，等.基于大模型的事件抽取技术及军事应用思考[J].网络安全与数据治理，2023,42(S1):163-168.
[7]周中元，刘小毅，李清伟，等. ChatGPT技术及其对军事安全影响[J].指挥信息系统与技术，2023,14(2):7-16.
[8]Liu H,Wang Y,Fan W,et al. Trustw-orthy AI:A computational perspective[J/OL]. 2021-07-12. http：//arxiv.org/abs/2107.06641.
[9]Wei J,Bosma M,Zhao V Y,et al. Finetuned language models are zero-shot learners[J/OL]. 2021-09-03.http：//arxiv.org/abs/2109.01652.
[10]岳增营，叶霞，刘睿珩.基于语言模型的预训练技术研究综述[J].中文信息学报，2021,35(9):15-29.
[11]Jhong K Y. Evaluating artificial intelligence for operations in the information environment[J]. Defense Science Review,2023,12(4):78-92.
[12]Liu P,Yuan W,Fu J,et al. Pre-train,prompt,and predict:A systematic survey of prompting methods in natural language processing[J/OL]. 2021-07-29. http：//arxiv.org/abs/2107.13586.
[13]Khoshnoodi M,Jain V,GAO M,et al. A comprehensive survey of accelerated generation techniques in large language models[J/OL]. 2024-05-15. http：//arxiv. org/abs/2405.13019.
[14]Christiano P, Leike J, Brown T B, et al. Deep reinforcement learning from human preferences[J/OL].2023-02-17. http：//arxiv.org/abs/1706.03741.
[15]Schulman J, Wolski F, Dhariwal P, et al. Proximal policy optimization algorithms[J/OL]. 2017-07-20.http：//arxiv.org/abs/1707.06347.
[16]李铮，方琼，刘津玮，等. 2024年美军无人装备领域发展分析[J].战术导弹技术，2025(2):1-12.
[17]张龙，雷震，冯轩铭，等.军事大模型：应用分析、关键技术和评估体系框架[C].第六届体系工程学术会议论文集——体系工程与高质量发展.长沙：国防科技大学系统工程学院，2024:1149-1162.
[18]赵日，赵鹏飞，程运江，等.人工智能技术在反舰作战中的应用研究[J].战术导弹技术，2019(5):86-91.
[19]潘长鹏，韩玉龙，庄益夫.舰载无人机编队协同对海突击作战效能评估指标体系研究[J].战术导弹技术，2019(2):25-32.
[20]张龙，王数，雷震，等. AIGC军事大模型评估体系框架研究[J].战术导弹技术，2025(1):42-52.
Basic Information:
DOI：10.16358/j.issn.1009-1300.20250059
China Classification Code:TP18;E91
Citation Information:
[1]Lei Zhen,Li Liwei,Zhang Long ,et al.Requirement analysis and development process for militarized application of large-scale models[J].Tactical Missile Technology,2026,No.235(01):49-60.DOI:10.16358/j.issn.1009-1300.20250059.