This project addresses two major challenges in today’s educational technology: how to build better smart education models and how to design integrated service systems that support them. Our work focuses on five main areas: (1) designing effective structures for models that can process different types of data (such as text, images, and video), (2) expanding what these models know by using knowledge networks called “knowledge forests,” (3) making sure the models reflect real teaching practices, (4) enabling AI systems to work together and learn from one another, and (5) creating strong methods to evaluate how well smart education services perform. We’re also introducing exciting new ideas, such as a “1+M” model structure and a “1+M+N” service system, which offer more flexibility and scalability. One of our most innovative breakthroughs is a new way of building large education models called the “tower construction” method. Unlike traditional fine-tuning techniques, this approach unlocks new capabilities in the models, creates new opportunities for research, and helps accelerate progress in multimodal artificial intelligence—ultimately making educational tools smarter, more adaptive, and more effective for learners and teachers alike.