The microbiome as a systems-level regulator of immune, metabolic, neural, and endocrine signaling in cancer.

The effectiveness of theory-based instruction has become an important issue in higher education, while art history is a foundational theory course in art education. However, due to the complexity of its theoretical framework, the teaching of art history faces numerous practical challenges. On one hand, the course content is highly abstract and vast, encompassing multiple theoretical dimensions such as iconography, formalism, aesthetic philosophy, and social art history. It requires students to not only master extensive historical background, artists' styles, and ideological systems, but also to possess strong interdisciplinary integration and abstract comprehension skills (). On the other hand, resources for art history education in universities are relatively scarce. Common issues such as insufficient faculty, lack of teaching aids, and limited personalized course support hinder students, especially those in resource-limited areas, from accessing high-quality and personalized art history education resources (). Traditional teaching in art history often relies on lectures and textbook reading; although this approach provides systematic content delivery, it depends heavily on teacher-led instruction and offers limited personalized support. As a result, students with different levels of prior knowledge may receive insufficient individualized support, which can limit deeper understanding and knowledge integration (). Consequently, many students generally struggle with understanding abstract theories, memorizing large amounts of information, and constructing systematic knowledge structures, which in turn affects their performance in art history learning and their ability to apply theoretical knowledge (). With the rapid development of artificial intelligence (AI) technology, particularly advances in large language models (LLMs), the application of LLMs in education has become an active research area (; ). Prior benchmark studies using examination-style questions have shown that LLMs can achieve competitive performance in structured educational tasks, including construction management (,) and medical education (), while prompting strategies such as self-consistency () and tree-of-thought () prompting may further improve reasoning performance (; ). Beyond benchmark evaluations, LLMs have also been tested in teaching and learning settings across engineering, medicine, and physics. For example, research by shows that the introduction of LLMs in university engineering education for assisted learning significantly enhances students' learning efficiency. demonstrate that the introduction of LLMs can effectively improve performance in diabetes-related exams and the efficiency of primary care physician training, with ChatGPT-4.0 () showing outstanding performance in both Chinese and English exams, suggesting potential value for medical education and clinical decision support. indicate that LLMs can provide timely feedback and explain domain knowledge in K-12 physics laboratory teaching, serving as a tool to enhance teaching efficiency and reduce teacher workload. Although these studies indicate the potential of LLMs for addressing complex teaching content and resource limitations, systematic evidence remains limited in humanities disciplines, especially art history.