mlsanityData to 7 Oct 2026

Adaptive Machine Unlearning

arXiv 2106.04378 ↗First posted 8 Jun 2021cs.LG stat.MLcited by 164

On the map: LLM safety and security → Machine unlearning

Abstract (arXiv)

Data deletion algorithms aim to remove the influence of deleted data points from trained models at a cheaper computational cost than fully retraining those models. However, for sequences of deletions, most prior work in the non-convex setting gives valid guarantees only for sequences that are chosen independently of the models that are published. If people choose to delete their data as a function of the published models (because they don't like what the models reveal about them, for example), then the update sequence is adaptive. In this paper, we give a general reduction from deletion guarantees against adaptive sequences to deletion guarantees against non-adaptive sequences, using differential privacy and its connection to max information. Combined with ideas from prior work which give guarantees for non-adaptive deletion sequences, this leads to extremely flexible algorithms able to handle arbitrary model classes and training methodologies, giving strong provable deletion guarantees for adaptive deletion sequences. We show in theory how prior work for non-convex models fails against adaptive deletion sequences, and use this intuition to design a practical attack against the SISA algorithm of Bourtoule et al. [2021] on CIFAR-10, MNIST, Fashion-MNIST.

Key facts. Cited by 164 papers in the map (data through 7 Oct 2026). Placed in Machine unlearning, one of 12 directions in LLM safety and security. Builds on 20 of its 43 references that are in the map. First posted on arXiv 8 Jun 2021.

Builds on

20 of 43 references.

Closest ancestors · 1

Same direction, newest first

  1. 2021Remember What You Want to Forget: Algorithms for Machine Unlearning4 Mar

Other references · 19

  1. 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,283
  2. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,221
  3. 2019Machine Unlearningcited by 1,020
  4. 2019Certified Data Removal from Machine Learning Modelscited by 512
  5. 2019Making AI Forget You: Data Deletion in Machine Learningcited by 447
  6. 2020Descent-to-Delete: Gradient-Based Methods for Machine Unlearningcited by 303
  7. 2018Scalable Private Learning with PATEcited by 266
  8. 2020Forgetting Outside the Box: Scrubbing Deep Networks of Information Accessible from Input-Output Observationscited by 171
  9. 2020Mixed-Privacy Forgetting in Deep Networkscited by 158
  10. 2020Tempered Sigmoid Activations for Deep Learning with Differential Privacycited by 121
  11. 2014Preserving Statistical Validity in Adaptive Data Analysiscited by 102
  12. 2015Algorithmic Stability for Adaptive Data Analysiscited by 88
7 more
  1. 2015Generalization in Adaptive Data Analysis and Holdout Reusecited by 76
  2. 2018Algorithms that Remember: Model Inversion Attacks and Data Protection Lawcited by 52
  3. 2018Privacy-preserving Predictioncited by 48
  4. 2019A New Analysis of Differential Privacy's Generalization Guaranteescited by 36
  5. 2016Max-Information, Differential Privacy, and Post-Selection Hypothesis Testingcited by 25
  6. 2018Mitigating Bias in Adaptive Data Gathering via Differential Privacycited by 21
  7. 2020Adversarially Robust Streaming Algorithms via Differential Privacycited by 21

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 7; 2022: 18; 2023: 34; 2024: 52; 2025: 34; 2026: 19.

164 papers.

Sort:

By area:

  1. 2023Detecting Pretraining Data from Large Language Modelscited by 375
  2. 2022A Survey of Machine Unlearningcited by 298
  3. 2021Unrolling SGD: Understanding Factors Influencing Machine Unlearningcited by 274
  4. 2024MUSE: Machine Unlearning Six-Way Evaluation for Language Modelscited by 251
  5. 2021Fast Yet Effective Machine Unlearningcited by 209
  6. 2024Machine Unlearning of Pre-trained Large Language Modelscited by 118
  7. 2024Guardrail Baselines for Unlearning in LLMscited by 90
  8. 2023Machine Unlearning: Solutions and Challengescited by 85
  9. 2024Inexact Unlearning Needs More Careful Evaluations to Avoid a False Sense of Privacycited by 78
  10. 2023DEPN: Detecting and Editing Privacy Neurons in Pretrained Language Modelscited by 75
  11. 2022Forget Unlearning: Towards True Data-Deletion in Machine Learningcited by 58
  12. 2024Position: LLM Unlearning Benchmarks are Weak Measures of Progresscited by 58
38 more
  1. 2024Langevin Unlearning: A New Perspective of Noisy Gradient Descent for Machine Unlearningcited by 48
  2. 2024Are we making progress in unlearning? Findings from the first NeurIPS unlearning competitioncited by 48
  3. 2021Hard to Forget: Poisoning Attacks on Certified Machine Unlearningcited by 47
  4. 2023Exploring the Landscape of Machine Unlearning: A Comprehensive Survey and Taxonomycited by 44
  5. 2023GNNDelete: A General Strategy for Unlearning in Graph Neural Networkscited by 43
  6. 2022Evaluating Machine Unlearning via Epistemic Uncertaintycited by 39
  7. 2022Algorithms that Approximate Data Removal: New Results and Limitationscited by 39
  8. 2022Hidden Poison: Machine Unlearning Enables Camouflaged Poisoning Attackscited by 37
  9. 2024Threats, Attacks, and Defenses in Machine Unlearning: A Surveycited by 34
  10. 2023Backdoor Attacks and Defenses in Federated Learning: Survey, Challenges and Future Research Directionscited by 33
  11. 2024Fast Exact Unlearning for In-Context Learning Data for LLMscited by 33
  12. 2024Machine Unlearning: A Comprehensive Surveycited by 33
  13. 2021Interpretable Data-Based Explanations for Fairness Debuggingcited by 30
  14. 2023Inductive Graph Unlearningcited by 30
  15. 2022Learning with Recoverable Forgettingcited by 29
  16. 2024Unlearning as multi-task optimization: A normalized gradient difference approach with an adaptive learning ratecited by 29
  17. 2023SAFE: Machine Unlearning With Shard Graphscited by 28
  18. 2022Repairing Neural Networks by Leaving the Right Past Behindcited by 27
  19. 2023Towards Machine Unlearning Benchmarks: Forgetting the Personal Identities in Facial Recognition Systemscited by 27
  20. 2024Corrective Machine Unlearningcited by 26
  21. 2023On the Safety of Open-Sourced Large Language Models: Does Alignment Really Prevent Them From Being Misused?cited by 25
  22. 2024Towards Scalable Exact Machine Unlearning Using Parameter-Efficient Fine-Tuningcited by 25
  23. 2024Reconstruction Attacks on Machine Unlearning: Simple Models are Vulnerablecited by 24
  24. 2023MultiDelete for Multimodal Machine Unlearningcited by 20
  25. 2024The Utility and Complexity of in- and out-of-Distribution Machine Unlearningcited by 20
  26. 2024Generative Unlearning for Any Identitycited by 19
  27. 2024CLIPErase: Efficient Unlearning of Visual-Textual Associations in CLIPcited by 19
  28. 2023Machine Unlearning: A Surveycited by 18
  29. 2024Unlearning Traces the Influential Training Data of Language Modelscited by 18
  30. 2024Attribute-to-Delete: Machine Unlearning via Datamodel Matchingcited by 18
  31. 2022Few-Shot Unlearning by Model Inversioncited by 17
  32. 2024Tighter Privacy Auditing of DP-SGD in the Hidden State Threat Modelcited by 17
  33. 2022Efficient Attribute Unlearning: Towards Selective Removal of Input Attributes from Feature Representationscited by 16
  34. 2022SIFU: Sequential Informed Federated Unlearning for Efficient and Provable Client Unlearning in Federated Optimizationcited by 16
  35. 2023Training Data Protection with Compositional Diffusion Modelscited by 16
  36. 2023Reinforcement Unlearningcited by 16
  37. 2024MU-Bench: A Multitask Multimodal Benchmark for Machine Unlearningcited by 16
  38. 2023Feature Unlearning for Pre-trained GANs and VAEscited by 15

50 most cited of 164 All 164 (JSON)