mlsanityData to 7 Oct 2026

Improving Unlearning with Model Updates Probably Aligned with Gradients

arXiv 2511.02435 ↗First posted 4 Nov 2025cs.LGcited by 1

On the map: LLM safety and security → Machine unlearning

Abstract (arXiv)

We formulate the machine unlearning problem as a general constrained optimization problem. It unifies the first-order methods from the approximate machine unlearning literature. This paper then introduces the concept of feasible updates as the model's parameter update directions that help with unlearning while not degrading the utility of the initial model. Our design of feasible updates is based on masking, ie a careful selection of the model's parameters worth updating. It also takes into account the estimation noise of the gradients when processing each batch of data to offer a statistical guarantee to derive locally feasible updates. The technique can be plugged in, as an add-on, to any first-order approximate unlearning methods. Experiments with computer vision classifiers validate this approach.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Machine unlearning, one of 12 directions in LLM safety and security. Builds on 31 of its 42 references that are in the map. First posted on arXiv 4 Nov 2025.

Builds on

31 of 42 references.

Closest ancestors · 8

Same direction, newest first

  1. 2024Machine Unlearning: A Comprehensive Survey13 May
  2. 2024Inexact Unlearning Needs More Careful Evaluations to Avoid a False Sense of Privacy2 Mar
  3. 2023SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation19 Oct
  4. 2023Machine Unlearning: A Survey6 Jun
  5. 2023Exploring the Landscape of Machine Unlearning: A Comprehensive Survey and Taxonomy10 May
  6. 2023Model Sparsity Can Simplify Machine Unlearning11 Apr
  7. 2021Unrolling SGD: Understanding Factors Influencing Machine Unlearning27 Sep
  8. 2021Machine Unlearning of Features and Labels26 Aug

Other references · 23

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  3. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  4. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,221
  5. 2020Gradient Surgery for Multi-Task Learningcited by 1,210
  6. 2019Machine Unlearningcited by 1,020
  7. 2021Membership Inference Attacks From First Principlescited by 839
  8. 2019Eternal Sunshine of the Spotless Net: Selective Forgetting in Deep Networkscited by 599
  9. 2019Certified Data Removal from Machine Learning Modelscited by 512
  10. 2021Conflict-Averse Gradient Descent for Multi-task Learningcited by 441
  11. 2018Lipschitz regularity of deep neural networks: analysis and efficient estimationcited by 413
  12. 2020Label-Only Membership Inference Attackscited by 371
11 more
  1. 2021Membership Inference Attacks on Machine Learning: A Surveycited by 306
  2. 2020Amnesiac Machine Learningcited by 305
  3. 2019Efficient and Accurate Estimation of Lipschitz Constants for Deep Neural Networkscited by 290
  4. 2020Mixed-Privacy Forgetting in Deep Networkscited by 158
  5. 2020Learning explanations that are hard to varycited by 132
  6. 2022Recommendation Unlearningcited by 71
  7. 2023Scalable Membership Inference Attacks via Quantile Regressioncited by 61
  8. 2021SAND-mask: An Enhanced Gradient Masking Strategy for the Discovery of Invariances in Domain Generalizationcited by 50
  9. 2020Formalizing Data Deletion in the Context of the Right to be Forgottencited by 46
  10. 2025The European Union general data protection regulation: what it is and what it meanscited by 30
  11. 2022l-Leaks: Membership Inference Attacks with Logitscited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Behavioral Guarantees for Proxy-Based Unlearningcited by 0