【作 者】：
余越凡， 赵志群， 黄方慧
【关 键 词 】：
大模型； 多智能体； 会话机器人； 职业能力； 支架式教学
【栏 目】：
其他
【中文摘要】：
尽管目前生成式大模型技术快速发展，但基于大模型的职业教育会话机器人构建仍面临诸多挑战，包括传统设计难以支撑职业能力发展、大模型推理能力和领域知识限制教学决策、职业能力诊断方法科学性不足等。为此，研究基于“设计导向”职业教育思想、“行动导向”职业教学原则、“从初学者到专家”能力发展逻辑和支架式教学等理论基础，综合应用多智能体、思维链、检索增强生成等技术，构建了基于大模型的职业教育会话机器人及其应用模式。研究采用准实验研究法，以高职“电子电路与CAD制版”课程为教学内容开展教学实验。结果表明，基于大模型的职业教育会话机器人及其应用模式能够显著提升学生的第一、第二级别职业能力。此外，该会话机器人能够在控制响应延迟和API调用成本的同时，改善大模型推理能力和领域知识不足问题，并实现了有效的能力诊断。研究还进一步探讨了未来基于大模型的职业教育会话机器人的优化思路。
【英文摘要】：
Despite the rapid development of generative large language models (LLMs), the construction of LLM-based vocational education chatbots still confronts some challenges, including the difficulty of traditional designs to support the development of professional competence, the limitations of the reasoning ability and domain knowledge of LLMs to teaching decision-making, and inadequate scientificity of professional competence assessment methods. To address these challenges, based on the theoretical foundation of the "design-oriented" vocational education ideology, the "action-oriented" vocational teaching principles, the "from novice to expert" competence development logic, and scaffolding teaching, the study constructs an LLM-based vocational education chatbot and its application model by comprehensively utilizing multi-agent systems, chain-of-thought prompting, and retrieval-augmented generation technologies. A quasi-experimental study in a course titled "Electronic Circuits and CAD Plate Making" is conducted in higher vocational education. The results demonstrate that the LLM-based vocational education chatbot and its application model can significantly improve students' level 1 and level 2 professional competence. Furthermore, the chatbot can improve LLM reasoning capabilities and domain knowledge deficiencies while controlling response delay and API call cost, therefore achieve effective competence assessment. This study further discusses optimizing strategies of the LLM-based vocational education chatbot.