Gaussian certified unlearning in high dimensions: A hypothesis testing approach

ICLROral2026

Authors: Aaradhya Pandey, Arnab Auddy, Haolin Zou, Arian Maleki, Sanjeev Kulkarni
Affiliation: Princeton University
Venue: ICLR 2026
Track: Oral

TL;DR

We introduce the canonical dimension free notion of certifiability suitable to high dimensions and show its utility via a Newton based unlearning algorithm…

Opening excerpt from the authors’ abstract. source

Read the paper

OpenReview

← All ICLR 2026 Oral papers · Browse the whole archive