Incorporating RAG for Factual Hallucination Detection Modeling in Intelligent Educational system

Factual hallucination is one of the key factors affecting learning effectiveness and knowledge construction in intelligent education system, and reducing factual hallucination can help develop learners' critical thinking and accurate knowledge acquisition. How to effectively detect factual hallucination in intelligent education system is a challenging issue in educational technology research. Based on this, this study proposed a factual hallucination detection model that incorporates knowledge retrieval enhancement and fine-tuning of large language model. First, an educational knowledge vector base is constructed from the collected knowledge datasets in the educational domain; then, a factual illusion detection model is constructed by fusing the retrieval augmented generation(RAG ) technique and Low-Rank Adaptation of Large Language Models(LoRA) fine-tuning method to optimize the large language model. We conducted experiments on a real dataset in the education domain, and the results showed that this experimental model outperforms the existing models in terms of precision, recall, and F1 score for factual hallucination detection based on knowledge retrieval enhancement and model fine-tuning. Therefore, the experimental model in this paper is able to effectively detect factitious hallucination in intelligent education system, which lays the foundation for providing high-quality educational content.