CONSTRUCTION AND PRACTICE OF AN INTERACTIVE TUTORING SYSTEM BASED ON LARGE MODEL KNOWLEDGE MODEL

This study explores innovative applications of large language models (LLMs) in college physics teaching and constructs a tutoring system via an agent platform. By inte-grating a college physics knowledge base into the agent, we significantly enhanced its abi-lity to solve physics problems and successfully applied it to four core scenarios: pre-class preparation, post-class review, exercise tutoring, and physics model visualization. For pre-class preparation and post-class review, the agent extracts and organizes key knowledge points from textbooks. In solving exercises in the textbook, the agent achieves nearly 100% accuracy for all exercises in college physics textbooks, allowing students to inquire further about unclear steps. For visualizing physics concepts, it converts physical models into executable code to generate diagrams, reducing learning difficulty. Student feedback indicates that this approach improves learning efficiency, lowers the difficulty of college physics, enhances interest, and provides a replicable model for AI-empowered foundational education reform.