mlsanityData to 7 Oct 2026

Machine unlearning

1,043 papers6 this weekHigh confidence in this grouping: the papers form one clear topicHigh confidence in this grouping: the papers form one clear topic · About

Keywords: unlearning, machine, machine unlearning, knowledge, forgetting

Key facts. Machine unlearning: 1,043 papers on arXiv since 2021 (data through 7 Oct 2026). 394 of its 1,043 mapped papers (38%) cite “TOFU: A Task of Fictitious Unlearning for LLMs” (2024, arXiv 2401.06121). 354 of its 1,043 mapped papers (34%) cite “Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning” (2024, arXiv 2404.05868). 40 new papers 24 Sep–7 Oct.

What’s new

24 Sep – 7 Oct 2026

40 new papers, 24 Sep – 7 Oct; 6 this week.

  1. This week · 6
  2. Enabling Preference-driven Unlearning in Few-step Distilled Text-to-Image Diffusion Models
  3. Nullify: Null-Space Activation Steering for Training-Free LLM Unlearning
  4. Why Forget-Only Unlearning Needs Memorization
  5. Open-MMUnlearning: Unifying Methods and Evaluation for MLLM Unlearning
  6. How Learning Governs Unlearning across the Memorization-Generalization Spectrum
  7. Exact Unlearning via Quantized Sufficient Statistics
  8. Earlier · 34
  9. Your Unlearning Gives You Away: Identifying Erased Concepts in Diffusion Models
  10. No Concept Escapes the Audit: Auditing-Aware Unlearning for Verifiable Concept Erasure in Diffusion Models
  11. UnAct: Gradient-Free Unlearning via Targeted Activation Intervention
  12. Modeling Deletion Requests in Machine Unlearning
All 40
  1. Certified Mechanistic Edits: Behavioral Guarantees for Skill Removal and Preservation also fits: Vision-language and text-to-image safety
  2. On Unlearning for Time-series Forecasting
  3. From Mathematical to Executable Certificates for Machine Unlearning
  4. SIEVE: Selective attention-value Suppression for Vision-Language Models Unlearning
  5. Repurposing Obsolete Representations for Post-Deployment Adaptation
  6. Linguistic Loopholes in LLM Unlearning: From a 174-Language Benchmark to Coverage-Aware Unlearning
  7. LLM Persona Unlearning
  8. Learning What to Forget: Distributional Unlearning for LLM Representation Spaces
  9. Unlearning Deceptive Behaviors in LLMs with Contrastive Forget Sets
  10. Unmerge: Efficient Machine Unlearning via Task Arithmetic
  11. Storage Is Not Strategy: State-Conditioned Support Control for LLM Unlearning
  12. Weeding Out Bad Seeds: Initial-Noise-Robust Unlearning for Text-to-Image Diffusion Models
  13. UnlearningSoup: Is Repeated Tuning Necessary for Large Language Model Unlearning?
  14. Do LLMs Really Forget? Hidden-State Leakage in Model Unlearning and How to Fix it
  15. Data Unlearning via Inverse Distillation
  16. UNBIND: UNlearning By INference-time Directional Steering for Code LLMs
  17. Bounding Retraining Equivalence and the Deletion Floor in Materials Machine Unlearning
  18. eval-unlearn: Benchmarking unlearning in Text-to-Image Diffusion Models also fits: Vision-language and text-to-image safety
  19. TULIP: Targeted LLM Unlearning at Layers Identified Per-Input
  20. Causal Routing for Unlearning
  21. Making LLMs Truly Forget: Deep Unlearning by Searching, Selecting, and Severing Knowledge Paths
  22. You Can't Have It Both Ways: Concept Entanglement Limits Diffusion Model Unlearning
  23. Evaluating Machine Unlearning in ASR
  24. Trajectory Unlearning on LLM-based Agents
  25. Continual Data Unlearning in Diffusion Models via Transition-based Regularization
  26. PruneForget: Joint Unlearning and Pruning of Vision Models
  27. LLM Unlearning Evaluation with TRIAGE
  28. Open Vocabulary Domain Unlearning also fits: Vision-language and text-to-image safety
  29. Neuralyzing the Trace: Selective Representation-Level Unlearning with Contrastive Sparse Autoencoders
  30. Inference-Time Target Speaker Unlearning in LLM-Based Automatic Speech Recognition

Foundations

Most cited by papers here

  1. TOFU: A Task of Fictitious Unlearning for LLMs2024cited by 394 here
  2. Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning2024cited by 354 here
  3. Knowledge Unlearning for Mitigating Privacy Risks in Language Models2022cited by 299 here
  4. Who's Harry Potter? Approximate Unlearning in LLMs2023cited by 267 here
  5. Remember What You Want to Forget: Algorithms for Machine Unlearning2021cited by 252 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2021Remember What You Want to Forget: Algorithms for Machine Unlearningcited by 252 here
  2. 2021Unrolling SGD: Understanding Factors Influencing Machine Unlearningcited by 232 here
  3. 2021On the Necessity of Auditable Algorithmic Definitions for Machine Unlearningcited by 123 here
  4. 2021Fast Yet Effective Machine Unlearningcited by 160 here
  5. 2022Can Bad Teaching Induce Forgetting? Unlearning in Deep Networks using an Incompetent Teachercited by 154 here
  6. 2022A Survey of Machine Unlearningcited by 176 here
  7. 2023Fast Machine Unlearning Without Retraining Through Selective Synaptic Dampeningcited by 141 here
  8. 2023In-Context Unlearning: Language Models as Few Shot Unlearnerscited by 169 here
  9. 2024TOFU: A Task of Fictitious Unlearning for LLMscited by 394 here
  10. 2024Eight Methods to Evaluate Robust Unlearning in LLMscited by 114 here
  11. 2024The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearningcited by 265 here
  12. 2024Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearningcited by 354 here
  13. 2024MUSE: Machine Unlearning Six-Way Evaluation for Language Modelscited by 212 here
  14. 2024Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearningcited by 131 here
10 more it builds on, not in this chain
  1. 2021Machine Unlearning of Features and Labelscited by 177 here
  2. 2022Continual Learning and Private Unlearningcited by 118 here
  3. 2022Knowledge Unlearning for Mitigating Privacy Risks in Language Modelscited by 299 here
  4. 2023Towards Unbounded Machine Unlearningcited by 209 here
  5. 2023Model Sparsity Can Simplify Machine Unlearningcited by 170 here
  6. 2023Who's Harry Potter? Approximate Unlearning in LLMscited by 267 here
  7. 2023Large Language Model Unlearningcited by 206 here
  8. 2023SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generationcited by 212 here
  9. 2023Unlearn What You Want to Forget: Efficient Unlearning for LLMscited by 154 here
  10. 2024Rethinking Machine Unlearning for Large Language Modelscited by 183 here

The 24 most cited of 1,147.

Browse

30 of 1,043, ranked 6–35

  1. 6Unrolling SGD: Understanding Factors Influencing Machine Unlearning2021232
  2. 7MUSE: Machine Unlearning Six-Way Evaluation for Language Models2024212
  3. 8SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation2023212
  4. 9Towards Unbounded Machine Unlearning2023209
  5. 10Large Language Model Unlearning2023206
  6. 11Rethinking Machine Unlearning for Large Language Models2024183
  7. 12Machine Unlearning of Features and Labels2021177
  8. 13A Survey of Machine Unlearning2022176
  9. 14Model Sparsity Can Simplify Machine Unlearning2023170
  10. 15In-Context Unlearning: Language Models as Few Shot Unlearners2023169
All 30
  1. 16Fast Yet Effective Machine Unlearning2021160
  2. 17Unlearn What You Want to Forget: Efficient Unlearning for LLMs2023154
  3. 18Can Bad Teaching Induce Forgetting? Unlearning in Deep Networks using an Incompetent Teacher2022154
  4. 19Fast Machine Unlearning Without Retraining Through Selective Synaptic Dampening2023141
  5. 20Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearning2024131
  6. 21On the Necessity of Auditable Algorithmic Definitions for Machine Unlearning2021123
  7. 22Continual Learning and Private Unlearning2022118
  8. 23Eight Methods to Evaluate Robust Unlearning in LLMs2024114
  9. 24Adaptive Machine Unlearning2021113
  10. 25Boundary Unlearning2023112
  11. 26Zero-Shot Machine Unlearning2022102
  12. 27Large Language Model Unlearning via Embedding-Corrupted Prompts2024101
  13. 28Machine Unlearning via Algorithmic Stability2021101
  14. 29Machine Unlearning of Pre-trained Large Language Models202493
  15. 30Towards Safer Large Language Models through Machine Unlearning202491
  16. 31SOUL: Unlocking the Power of Second-Order Optimization for LLM Unlearning202489
  17. 32Guardrail Baselines for Unlearning in LLMs202479
  18. 33KGA: A General Machine Unlearning Framework Based on Knowledge Gap Alignment202378
  19. 34An Adversarial Perspective on Machine Unlearning for AI Safety202476
  20. 35Reversing the Forget-Retain Objectives: An Efficient LLM Unlearning Framework from Logit Difference202475