Scholarlink

This paper introduces CoPaa, an LLM-powered system for automated curriculum outline generation, addressing limitations of existing methods in efficiency and content quality. CoPaa integrates five modules: a Content Parsing Engine for extracting textbook structures, an Instruction Chain Generator for task decomposition, a Previous Step Description Generator for maintaining instructional coherence, a Parallel Output Engine for efficient content generation using LangGraph, and a Controllable Calibration Optimizer for human-in-the-loop refinement. To evaluate performance, we propose CoEval, a multi-dimensional framework. Experimental results show CoPaa achieves an average score of 8.52/10, with expert validation and ablation studies confirming the effectiveness of each module. CoPaa provides a scalable, pedagogically sound solution for curriculum design.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

CoPaa: Curriculum Outline Process Automatic Agent Based on LangGraph

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Jiarui Zhang,
  • Qiming Zhang,
  • Yuqing He,
  • Yanbin Zhang

摘要

This paper introduces CoPaa, an LLM-powered system for automated curriculum outline generation, addressing limitations of existing methods in efficiency and content quality. CoPaa integrates five modules: a Content Parsing Engine for extracting textbook structures, an Instruction Chain Generator for task decomposition, a Previous Step Description Generator for maintaining instructional coherence, a Parallel Output Engine for efficient content generation using LangGraph, and a Controllable Calibration Optimizer for human-in-the-loop refinement. To evaluate performance, we propose CoEval, a multi-dimensional framework. Experimental results show CoPaa achieves an average score of 8.52/10, with expert validation and ablation studies confirming the effectiveness of each module. CoPaa provides a scalable, pedagogically sound solution for curriculum design.