Large Language Models in Student Simulation: A Survey

Large language models (LLMs) are transforming educational research. Within its many usage, LLM-based simulations emerge as a powerful paradigm for modeling student cognition, behavior, and learning variability. In this survey, we systematically review recent advances in LLM-driven student simulation, examining how researchers adapt, fine-tune, or prompt models to emulate varying levels of student proficiency and engagement. In addition, we categorize existing studies based on their simulation objectives, model adaptation strategies, and validation methodologies. Our survey offers a technological overview to help educators and researchers understand the potential of LLMs to advance educational research and practice.