BPE: Exploring the Prompt Framework of Physics Exercises Generated from Bloom's Taxonomy in LLM

Haoran Luo, Jidong Weng, Xunuo Lu, Zengyi Yu

Abstract: This study explores the application of large language models (LLMs) in automating the generation of physics exercises. By integrating principles from Bloom's Taxonomy, the research develops a prompt framework named BPE (Bloom's Physics Exercises) to enhance the quality and relevance of automatically generated questions. The BPE framework addresses challenges such as controlling question difficulty, assessing cognitive depth, and managing computational aspects. Key components of the framework include ‘Action’, ‘Purpose’, ‘Example’, ‘Role’, and ‘Excluded Content’, which together guide the LLM to produce exercises that are pedagogically sound and tailored to educational needs. This approach aims to reduce teacher workload, diversify student learning experiences, and improve the overall educational efficacy of AI-assisted learning environments. Our contributions are available at https://github.com/yuzengyi/BPE.