Beyond Sample-Level Forgetting: Improving Reliability in Multimodal Unlearning
Abstract
Multimodal unlearning seeks to remove specific data from pretrained multimodal models while preserving reliability and locality. This work uses decoupled knowledge components, Multimodal Variational Inference, and contrastive semantic editing to improve refined forgetting under privacy- and copyright-sensitive scenarios.
Type
Publication
In Proceedings of the 43rd International Conference on Machine Learning (ICML 2026)
Jianzhou Wang
M.E. student
Yirui Wu
Young Professor, CCF Senior Member
My research interests include Computer Vision, Artifical Intelligence, Multimedia Computing and Intelligent Water Conservancy.
Lixin Yuan
Lecturer
