Developing Teacher-Like AGI (Artificial General Intelligence) Robots for K-12 Math Education Using Reinforcement Learning From Reverse-Generated Data

• In Event: Virtual Poster Session
  In Virtual iPresentation Only: SIG-Advanced Technologies for Learning Virtual Poster Session

Abstract

Emerging technologies such as deep neural networks and large language models (LLMs) offer significant potential to enhance learning experiences, but single-function automated tools are inadequate for meeting the comprehensive demands of modern education. This paper proposes a framework leveraging Reinforcement Learning from Reverse Generated Data (RLRGD) to develop teacher-like tutoring robots for K-12 math education. By pre-training LLMs on authentic mathematical datasets and employing RLRGD to fine-tune models to emulate human teacher outputs, our approach aims to create intelligent systems capable of providing educationally meaningful and facilitative support across various mathematical topics. This framework addresses the limitations of current educational technologies, offering robust and adaptive tools to improve learning outcomes for K-12 students.

Authors

• Fan Zhang, University of Florida

• Wanli Xing, University of Florida