报告题目：Knowledge-based Enhancement of Foundation Models

报告时间：2025年12月11日（周四） 16:00

报告地点：华南理工大学国际校区 F3-b113

邀请人：余翔宇副教授

主持人：袁红老师

欢迎广大师生参加！

摘要：

Foundation models, including Large Language Models (LLMs) and Vision-language Models (VLMs), demonstrate exceptional performance across various tasks but face challenges in integrating knowledge efficiently and effectively. These challenges stem primarily from the need for specialized domain knowledge, such as that in medical fields, and the continual evolution of real-world knowledge which can render the information within model parameters obsolete. This talk will present recent advancements from my research group aimed at enhancing foundation models with knowledge. We treat knowledge at three levels: embeddings, graphs, and data. I will discuss strategies for effectively retrieving task-relevant knowledge, integrating this knowledge to address complex questions, and instilling multi-modal knowledge to boost model performance and explainability.

个人简历：

Dr. Ninghao Liu is an assistant professor in the Department of Computing at The Hong Kong Polytechnic University. He was an assistant professor in the School of Computing at the University of Georgia. His research work spans Explainable AI (XAI), Large Foundation Models, and Graph Mining. He has published papers at recognized venues such as NeurIPS, ICLR, KDD, WWW, ICML, CIKM, etc. His work won the Distinguished Paper Award in AMIA 2024, Outstanding Paper Award in ICML 2022, and the Best Paper Award Candidate in ICDM 2019. His work was supported by National Science Foundation (NSF), Institute of Education Sciences (IES) and Tencent America. He also received Google Research Scholar award in 2025.