REAT：面向多轮数学辅导的反思性经验增强框架
原标题：REAT: A Reflective Experience-Augmented Tutoring Framework for Multi-turn Mathematical Instruction
AI 摘要
该论文提出REAT（反思性经验增强辅导）框架，通过多智能体Observer-Critic-Mentor（OCM）蒸馏流程，将历史辅导对话提炼为结构化、与问题无关的教学经验，并在实时辅导中通过状态感知检索模块注入这些经验，以根据学生认知状态提供自适应支架。实验表明，该框架显著优于仅提示和SFT基线，尤其在复杂、低分辅导场景中提升明显，且蒸馏经验在不同模型架构和数学数据集上具有稳健的泛化能力。
正文节选
REAT: A Reflective Experience-Augmented Tutoring Framework for Multi-turn Mathematical Instruction Abstract Current Large Language Models (LLMs) excel at solving complex mathematical problems, yet this proficiency does not inherently translate into effective tutoring. While advanced LLM tutors may leverage multi-agent frameworks or fine-tuning, most still lack a mechanism to systematically accumulate and reuse pedagogical experience over time, limiting their adaptability to diverse student needs