AI RESEARCH

Fast Unlearning at Scale via Margin Self-Correction

arXiv CS.LG

ArXi:2606.02920v1 Announce Type: new Language-model unlearning updates a trained model to behave as if it had not seen selected