Optimization and Application of Large Language Model in Higher Education Question Answering System

With the rapid development of large language model technology, its application in diverse domains has expanded significantly, and higher education is no exception. This study investigates the optimization and application of large language models in higher education knowledge question-answering systems. First, it reviews the theoretical foundations of large language models and domain-oriented question-answering systems, clarifying their core architectures, training paradigms, and typical deployment modes in educational contexts. Second, it examines strategies to enhance system performance from three key dimensions: data and model optimization, integration of structured and unstructured knowledge resources, and the design of effective human-computer interaction mechanisms. On this basis, the study elaborates on innovative applications across disciplines, including intelligent tutoring, personalized learning support, formative assessment, academic research assistance, and institutional knowledge management. Furthermore, it analyzes major challenges such as data quality, model bias, interpretability, privacy protection, and alignment with pedagogical goals, and summarizes corresponding coping strategies and governance frameworks. Finally, the paper discusses future development trends, including multimodal integration, adaptive learning analytics, and closer collaboration between educators and AI systems. The study aims to provide a systematic reference for the effective deployment of large language models in higher education knowledge question-answering systems and to promote the transformation and sustainable development of higher education teaching models.