Abstract: STEM education is an important pathway for China to promote the cultivation of innovative talent and a key support for enhancing students’ comprehensive competencies. Currently, constructing a multifunctional teaching-agent system based on a multi-agent collaboration mechanism—and deploying it as a platform to empower STEM education—has become a key driver in generating new forms of STEM education amid the push for digital transformation in education. It is thus significant to propose the “CGS-AI” framework and develop a multi-agent “Intelligent Learning Companion” system designed to guide primary school students through STEM learning activities. Experimental results demonstrate that in STEM education supported by such multi-agent systems, employing heuristic dialogue with the “Intelligent Learning Companion” as the primary mode of collaborative learning leads to positive student outcomes across three dimensions: deeper conceptual understanding, enhanced practical skills, and strengthened learning motivation. The findings not only offer empirical support for the application of agent technology in education but also shed light on the underlying mechanisms of technology-driven paradigm shift in educational practice.

Key words: new form of STEM education, teaching agent, multi-agent system, artificial intelligence, intelligent learning companion