Name: Prof. Xiangnan He
Affiliation: University of Science and Technology of China, China
Speech Title: Large Model Personalization: Frontiers and Outlook
Biography:
Dr. Xiangnan He is a Professor and Associate Dean of the School of Artificial Intelligence at the University of Science and Technology of China (USTC), a recipient of the National Outstanding Youth Fund. His research interests span information recommendation and mining, large models, and artificial general intelligence. He has published over 100 papers in leading international conferences (including SIGIR, WWW, and KDD) and top-tier journals (such as IEEE TKDE and ACM TOIS), and has been recognized as an Elsevier Highly Cited Chinese Researcher, with Google Scholar citation over 70,000 times. His honors include the Best Paper Award at SIGIR and ICLR, the Frontier Science Award at the International Congress of Basic Science, the Alibaba DAMO Academy Young Scientist Award, the CCF Young Scientist Award, and the Asian Young Scientist Award, etc. He serves as Associate Editor for several prestigious journals, including IEEE TKDE, IEEE TBD, and ACM TOIS, and has led multiple national-level research projects, such as key programs of the NSFC and National Key R&D Programs of the Ministry of Science and Technology.
Abstract:
With the breakthroughs of large language models in general intelligence, they have demonstrated powerful cognitive capabilities in understanding, generation, decision-making and other dimensions. However, general-purpose large models still face substantial challenges in satisfying individualized demands and scenario-specific tasks. Enabling large models to "understand individuals" and realize deep adaptation to users, organizations and even industries has become a key issue in promoting the practical application of artificial intelligence. This talk will review the core progress and cutting-edge challenges of large models in the field of personalization, focusing on the key technologies of personalized large models, including such innovative ideas as efficient fine-tuning for personalized data, dynamic modeling and retrieval mechanisms for long-term memory, agentic frameworks for complex tasks, and controllable model editing for knowledge updating. Finally, the future directions of personalized large models are prospected, including reinforcement learning-driven adaptive optimization, cloud-edge collaborative privacy-preserving computing, and continuously evolving multi-agent collaboration systems.