Two recent works were accepted to ICLR 2026: Paper 1 Paper 2 I am excited to share results spanning robust unlearning and practical deployment perspectives.