Unlearning Data at Scale
An ICML 2026 tutorial presented by Vinith M. Suriyakumar and Gautam Kamath, with contributions from Ayush Sekhari and Ashia Wilson.
Machine unlearning studies how to efficiently and effectively remove training data from a model that has already been trained. In this tutorial, we cover motivations, definitions, algorithms (with and without provable guarantees), and future directions for machine unlearning.