Machine unlearning has attracted increasing attention in the field of trustworthy AI. The goal of this process is to make an AI model "forget" targeted information with only slight modifications, but what do we truly mean by "forgetting"?
This talk first introduces machine unlearning in the context of privacy, which directly corresponds to the "Right to be Forgotten" from GDPR. We will then dive into our empirical studies on unlearning for Large Language Models (LLMs), which highlight critical pitfalls in current empirical evaluation methods. Finally, we will cover our recent attempt to formalize knowledge unlearning for LLMs, a concept that is significantly different from unlearning for privacy.

Eli Chien is an Assistant Professor at National Taiwan University (Electrical Engineering) with the Yushan Young Fellow. He was a visiting researcher at Google Research (hosted by Peter Kairouz) and a Postdoctoral Fellow at the Georgia Institute of Technology, working with Professor Pan Li. He obtained his Ph.D. from the University of Illinois, Urbana-Champaign, advised by Professor Olgica Milenkovic. His current research focuses on privacy in machine learning, including machine unlearning and differential privacy, as well as their applications to graph machine learning. His previous research centered on designing better graph neural networks with theoretical guarantees. His work has been primarily published in top-tier machine learning, data mining, and information theory venues, including ICML, NeurIPS, ICLR, Transactions on IT, KDD, TheWebConf, AISTATS, AAAI and more.