Retrieval-Augmented Generation for Smart Classroom Applications

Our study proposes an intelligent teaching method based on RAG, which dynamically integrates authoritative knowledge bases with language models to address the limitations of large models in factual accuracy and domain-specific expertise. The experiments employ three evaluation metrics to validate the effectiveness of RAG in educational scenarios. The results demonstrate that the RAG-enhanced model (54.8% accuracy) significantly outperforms baseline models (Llama2-7B at 38.8%), and ablation studies confirm the critical role of the retrieval module and factual verification in performance improvement. The system supports personalized teaching resource generation, offering a new paradigm for AI-driven education.