Euroswarms/CommandNet-Small

CommandNet数据集是一个包含100,000个合成战术训练场景的综合数据集，专为训练AI模型在叛乱和混合战争环境中作为高级小单位（班/排级别，8-35人）战术教官而设计。该数据集将军事训练演习数据转化为结构化的、富含推理的教学内容，适用于微调大型语言模型，以培养战术决策、认知建模和任务指挥能力。数据集覆盖时间跨度为2000-2030年，主要领域包括叛乱行动、反叛乱、混合战争和常规战争，输出格式为ShareGPT JSONL。每个场景包含12个详细部分：情况（班/排级别）、装备可用性、条件检查（任务指挥）、团队态势感知发展、认知建模与推理深度、压力响应与心理韧性、三个阶段（接触移动、接触演练执行、决策树分支）、战术报告协议、目标安全以及教官笔记与深度推理。此外，数据集还涵盖10个不同的理论框架、5个作战战区和9种地形类型，提供全面的战术训练内容。

The CommandNet dataset is a comprehensive collection of 100,000 synthetic tactical training scenarios designed for training AI models to serve as advanced small unit (squad/platoon level, 8-35 personnel) tactical instructors in insurgency and hybrid warfare environments. It transforms military training exercise data into structured, reasoning-rich instructional content suitable for fine-tuning large language models on tactical decision-making, cognitive modeling, and mission command competencies. The dataset spans the time period from 2000 to 2030, with primary domains including insurgency operations, counterinsurgency, hybrid warfare, and conventional warfare, and is formatted as ShareGPT JSONL. Each scenario consists of 12 detailed sections: Situation (platoon/squad level), Equipment Availability, Conditions Check (Mission Command), Team Situational Awareness Development, Cognitive Modeling & Reasoning Depth, Stress Response & Mental Resilience, three phases (Movement to Contact, Contact Drill Execution, Decision Tree Branches), Tactical Reporting Protocols, Objective Securement, and Instructor Notes & Deep Reasoning. Additionally, the dataset covers 10 distinct doctrinal frameworks, 5 theaters of operation, and 9 terrain types, providing comprehensive tactical training content.