【摘要】 人工智能技术不断突破，教育大模型成为人工智能时代促进教育变革创新的关键支撑技术。研发中国版教育大模型能更好实现自主可控和安全可信，相关研究需要基于国产算力重新设计系统架构，并在幻觉应对、推理逻辑优化、多模态协同、安全可控、生态建设、效果评估等六方面推进技术创新。实践表明，教育大模型在辅助教师备课、创新生成式课堂和促进精准教学等方面应用成效较为明显，但仍存在教育适用性不足、研发模式不完善、教师智能素养欠缺、缺乏应用参照和指导等问题。为充分发挥大模型作用，助力教学创新发展，需要多方施策加大教育专属大模型支持力度，加快重点场景应用标杆和案例打造，制定教师智能素养专项提升计划，建立安全可控“政产学研用”协同的教育大模型应用生态。更多还原

【Abstract】 With continuous breakthroughs in artificial intelligence(AI) technology, educational large language models(LLMs) have emerged as a pivotal technology driving educational transformation and innovation in the AI era. Developing a Chinese version of educational LLMs is essential to achieve autonomous control, security, and trustworthiness. Related research necessitates redesigning system architectures based on domestic computing power and advancing technological innovations in six key areas: hallucination mitigation, reasoning logic optimization, multimodal collaboration, safety and controllability, ecosystem development, and effectiveness evaluation. Practical applications demonstrate that educational LLMs have significantly assisted teachers’ efficacy in preparing teaching materials, innovating generative classrooms, and enabling precise teaching model. However, challenges persist, including insufficient educational applicability, imperfect R&D models, a lack of intelligent literacy among teachers, and the absence of application references and guidelines. To fully release the potential development of large models in fostering teaching innovation, multi-pronged strategies are required: increasing support for education LLMs, accelerating the development of benchmark applications in key scenarios, implementing specialized programs to enhance teachers’ intelligent literacy, and establishing a collaborative “government-industry-academia-research-application” ecosystem for educational LLMs deployment.更多还原