cs.LGNov 4, 2026

Do unlearning algorithms learn the right features?

Authors: Polina Dolgova, Amartya Sanyal, Sebastian U. Stich

Organizations: CISPA Helmholtz Center for Information Security · Saarland University · Department of Computer Science, University of Copenhagen

Abstract

Machine unlearning is commonly evaluated with standard post-deletion metrics that compare an unlearned model to one retrained on the remaining data. Such snapshot evaluations can conceal consequential differences in the features the two models encode. We prove that if an unlearned and a retrained model agree on their outputs but encode different features, a single gradient step on their prediction heads can break this agreement. We further show that deleting an arbitrarily small fraction of the data can cause retraining to select qualitatively different features, so matching retraining can require learning features that the full model never had. We then introduce a measure of feature recovery that separates missing from extraneous features. Across eight unlearning methods, models that perform well on standard metrics often fail to recover the retrained features. This mismatch predicts deviations from retraining under subsequent relearning and downstream adaptation, and directly correcting it systematically reduces these deviations.

Figures & tables

Appendix figures & tables19 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. An Illusion of Unlearning? Assessing Machine Unlearning Through Internal Representations

    Apr 9, 2026Yichen Gao, Altay Unal, Akshay Rangamani +1Machine UnlearningMachine Unlearning Evaluation

  2. We Should Distinguish Unlearning From Untraining

    Apr 9, 2026Eleni Triantafillou, Ahmed Imtiaz Humayun, Monica Ribero +3Machine UnlearningMachine Unlearning Evaluation

  3. Why Forget-Only Unlearning Needs Memorization

    Oct 7, 2026Luka Radić, Vikrant Singhal, Amartya SanyalMachine Unlearning