针对现有程序设计类课程分层教学中分层标准静态化、评价单一等问题，提出动态能力适配的“三阶四维”模型，具体阐述如何依托四维能力雷达图（逻辑结构、算法思维等）动态划分基础层—发展层—挑战层，结合知识深度与认知负荷设计分层路径，融合课堂任务卡与智能补偿系统构建双环反馈，建立动态赋权评价体系，通过实践教学数据说明该模型能显著提升学习者的代码规范性及工程能力，为新工科编程教学改革提供有效范式。

To address issues in existing hierarchical teaching models for programming courses, such as static stratification criteria and singular evaluation methods, this study proposes "a three-stage four-dimensional" model based on dynamic capability adaptation. Leveraging a four-dimensional competency radar chart(logical structure, algorithmic thinking, etc.), students are dynamically categorized into foundational, developmental, and challenge tiers. A hierarchical learning pathway is designed by integrating knowledge depth and cognitive load, while a dual-loop feedback mechanism is constructed through classroom task cards and an intelligent compensation system. A dynamically weighted evaluation framework is also established. Practical results demonstrate that this model significantly improves code standardization and engineering capabilities, offering an effective paradigm for programming education reform in emerging engineering disciplines.