系统工程与电子技术 ›› 2026, Vol. 48 ›› Issue (10): 3562-3572.doi: 10.12305/j.issn.1001-506X.2026.10.26

• 系统工程 • 上一篇    

军事智能体测评体系:理论框架与测评方法

梁起明1,2, 崔翛龙2, 盛萱竺1,2, 李永2, 高志强2   

  1. 1. 武警工程大学研究生大队,陕西 西安 710086
    2. 武警工程大学教育部重点实验室,陕西 西安 710086
  • 收稿日期:2025-07-18 出版日期:2026-10-25 发布日期:2026-09-30
  • 通讯作者: 崔翛龙
  • 作者简介:梁起明(1996—),男,博士研究生,主要研究方向为智能体、深度学习、军事理论
    盛萱竺(1999—),女,博士研究生,主要研究方向为智能体、大数据、数据标签
    李 永(1981—),男,教授,博士,主要研究方向为深度学习、模式识别、武警信息化
    高志强(1990—),男,副教授,博士,主要研究方向为智能体、大数据、人工智能
  • 基金资助:
    国家社会科学基金(2025-SKJJ-A-007)资助课题

Military agent evaluation system: theoretical framework and evaluation methods

Qiming Liang1,2, Xiaolong Cui2, Xuanzhu Sheng1,2, Yong Li2, Zhiqiang Gao2   

  1. 1. Graduate Student Brigade,Engineering University of PAP,Xi’an 710086,China
    2. Key Laboratory of Ministry of Education,Engineering University of PAP,Xi’an 710086,China
  • Received:2025-07-18 Online:2026-10-25 Published:2026-09-30
  • Contact: Xiaolong Cui

摘要:

针对军事智能体研究理论滞后于工程实践的问题,梳理军事智能体的基础理论,提出军事智能体的基本概念,分析军事智能体的基本特征、基础能力和面临的主要矛盾;总结智能体的测评方法,分析现有方法在军事智能体测评方面的局限性。在此基础上,提出军事智能体组件效能、系统成熟度和典型任务三大指标,以及人工主导测评和任务导向测评相结合的军事智能体测评体系,并在艾武架构智能体上进行了验证实验,为军事智能体理论研究和规范化发展提供了参考。

关键词: 军事智能体, 测评基准, 大模型, 多维度评估

Abstract:

In response to the fact that the research theory of military agents lags behind engineering practice, the basic theories of military agents are reviewed, the basic concepts of military agents are proposed, the basic characteristics, basic capabilities, and main contradictions faced by military agents are analyzed. The evaluation methods of agents are summarized, and the limitations of existing methods in military agent evaluation are analyzed. On this basis, three major indicators of military agents component effectiveness, system maturity, and typical tasks are proposed, as well as a military agents evaluation system combining manual led evaluation and task oriented evaluation. Validation experiments are conducted on the AiWu architecture agents, providing reference for the theoretical research and standardized development of military agents.

Key words: military agent, evaluation benchmark, large model, multi-dimensional evaluation

中图分类号: