mlsanityData to 14 Oct 2026

A Comprehensive Survey of Machine Unlearning Techniques for Large Language Models

arXiv 2503.01854 ↗First posted 22 Feb 2025cs.CL cs.AIcited by 34

On the map: LLM safety and security → Machine unlearning

Abstract (arXiv)

This study investigates the machine unlearning techniques within the context of large language models (LLMs), referred to as textitLLM unlearning. LLM unlearning offers a principled approach to removing the influence of undesirable data (e.g., sensitive or illegal information) from LLMs, while preserving their overall utility without requiring full retraining. Despite growing research interest, there is no comprehensive survey that systematically organizes existing work and distills key insights; here, we aim to bridge this gap. We begin by introducing the definition and the paradigms of LLM unlearning, followed by a comprehensive taxonomy of existing unlearning studies. Next, we categorize current unlearning approaches, summarizing their strengths and limitations. Additionally, we review evaluation metrics and benchmarks, providing a structured overview of current assessment methodologies. Finally, we outline promising directions for future research, highlighting key challenges and opportunities in the field.

Key facts. Cited by 34 papers in the map (data through 14 Oct 2026). Placed in Machine unlearning, one of 12 directions in LLM safety and security. Builds on 42 of its 54 references that are in the map. First posted on arXiv 22 Feb 2025.

Builds on

42 of 54 references.

Closest ancestors · 29

Same direction, newest first

  1. 2024Benchmarking Vision Language Model Unlearning via Fictitious Facial Identity Dataset5 Nov
  2. 2024Extracting Unlearned Information from LLMs with Activation Steering4 Nov
  3. 2024Protecting Privacy in Multimodal Large Language Models with MLLMU-Bench29 Oct
  4. 2024CLEAR: Character Unlearning in Textual and Visual Modalities23 Oct
  5. 2024Catastrophic Failure of LLM Unlearning via Quantization21 Oct
  6. 2024Mechanistic Unlearning: Robust Knowledge Unlearning and Editing via Mechanistic Localization16 Oct
  7. 2024A Closer Look at Machine Unlearning for Large Language Models10 Oct
  8. 2024An Adversarial Perspective on Machine Unlearning for AI Safety26 Sep
  9. 2024On Effects of Steering Latent Representation for Large Language Model Unlearning12 Aug
  10. 2024Machine Unlearning in Generative AI: A Survey30 Jul
  11. 2024Unlearning or Obfuscating? Jogging the Memory of Unlearned LLMs via Benign Relearning19 Jun
  12. 2024Opt-Out: Investigating Entity-Level Unlearning for Large Language Models via Optimal Transport18 Jun
17 more
  1. 2024Intrinsic Test of Unlearning Using Parametric Knowledge Traces17 Jun
  2. 2024RWKU: Benchmarking Real-World Knowledge Unlearning for Large Language Models16 Jun
  3. 2024Reversing the Forget-Retain Objectives: An Efficient LLM Unlearning Framework from Logit Difference12 Jun
  4. 2024Large Language Model Unlearning via Embedding-Corrupted Prompts12 Jun
  5. 2024RKLD: Reverse KL-Divergence-based Knowledge Distillation for Unlearning Personal Information in Large Language Models4 Jun
  6. 2024Single Image Unlearning: Efficient Machine Unlearning in Multimodal Large Language Models21 May
  7. 2024Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning8 Apr
  8. 2024Machine Unlearning for Traditional Models and Large Language Models: A Short Survey1 Apr
  9. 2024Guardrail Baselines for Unlearning in LLMs5 Mar
  10. 2024Eight Methods to Evaluate Robust Unlearning in LLMs26 Feb
  11. 2024Machine Unlearning of Pre-trained Large Language Models23 Feb
  12. 2024Towards Safer Large Language Models through Machine Unlearning15 Feb
  13. 2024TOFU: A Task of Fictitious Unlearning for LLMs11 Jan
  14. 2023In-Context Unlearning: Language Models as Few Shot Unlearners11 Oct
  15. 2022Knowledge Unlearning for Mitigating Privacy Risks in Language Models4 Oct
  16. 2022Quark: Controllable Text Generation with Reinforced Unlearning26 May
  17. 2022Continual Learning and Private Unlearning24 Mar

Other references · 13

  1. 2017Proximal Policy Optimization Algorithmscited by 15,215
  2. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  3. 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,955
  4. 2023MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Modelscited by 1,652
  5. 2022Editing Models with Task Arithmeticcited by 1,120
  6. 2024Refusal in Language Models Is Mediated by a Single Directioncited by 1,032
  7. 2024The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearningcited by 509
  8. 2023DEPN: Detecting and Editing Privacy Neurons in Pretrained Language Modelscited by 79
  9. 2024From Theft to Bomb-Making: The Ripple Effect of Unlearning in Defending Against Jailbreak Attackscited by 29
  10. 2024Reference-free Hallucination Detection for Large Vision-Language Modelscited by 23
  11. 2024Ethos: Rectifying Language Models in Orthogonal Parameter Spacecited by 20
  12. 2025Internal Activation Revision: Safeguarding Vision Language Models Without Parameter Updatecited by 16
1 more
  1. 2024Towards Transfer Unlearning: Empirical Evidence of Cross-Domain Bias Mitigationcited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 15; 2026: 19.

34 papers.

Sort:

By area:

  1. 2025A Survey of Self-Evolving Agents: What, When, How, and Where to Evolve on the Path to Artificial Super Intelligencecited by 170
  2. 2025A Comprehensive Survey in LLM(-Agent) Full Stack Safety: Data, Training and Deploymentcited by 116
  3. 2025Model Unlearning via Sparse Autoencoder Subspace Guided Projectionscited by 18
  4. 2025LLM Unlearning with LLM Beliefscited by 17
  5. 2025Do LLMs Really Forget? Evaluating Unlearning with Knowledge Correlation and Confidence Awarenesscited by 14
  6. 2025SAUCE: Selective Concept Unlearning in Vision-Language Models with Sparse Autoencoderscited by 13
  7. 2025CRISP: Persistent Concept Unlearning via Sparse Autoencoderscited by 11
  8. 2025LUNE: Efficient LLM Unlearning via LoRA Fine-Tuning with Negative Examplescited by 8
  9. 2026Agentic Unlearning: When LLM Agent Meets Machine Unlearningcited by 8
  10. 2026From Weights to Activations: Is Steering the Next Frontier of Adaptation?cited by 8
  11. 2026Multimodal Unlearning Across Vision, Language, Video, and Audio: Survey of Methods, Datasets, and Benchmarkscited by 7
  12. 2025Beyond Chunks and Graphs: Retrieval-Augmented Generation through Triplet-Driven Thinkingcited by 6
22 more
  1. 2025UniErase: Towards Balanced and Precise Unlearning in Language Modelscited by 5
  2. 2025DUSK: Do Not Unlearn Shared Knowledgecited by 4
  3. 2026Reinforcement Unlearning via Group Relative Policy Optimizationcited by 4
  4. 2025LLM Unlearning Should Be Form-Independentcited by 3
  5. 2026Visual-Noise Guided In-Context Distillation for Multimodal Large Language Model Unlearningcited by 3
  6. 2025Feature-Selective Representation Misdirection for Machine Unlearningcited by 2
  7. 2026Forgetting Only What Matters: Layer-Selective Unlearning toward Robust LLMscited by 2
  8. 2025Beyond Sharp Minima: Robust LLM Unlearning via Feedback-Guided Multi-Point Optimizationcited by 1
  9. 2026Steerability of Instrumental-Convergence Tendencies in LLMscited by 1
  10. 2026ForgetMark: Stealthy Fingerprint Embedding via Targeted Unlearning in Language Modelscited by 1
  11. 2026Anatomy of Unlearning: The Dual Impact of Fact Salience and Model Fine-Tuningcited by 1
  12. 2026Trustworthy Agent Network: Trust in Agent Networks Must Be Baked In, Not Bolted Oncited by 1
  13. 2026Beyond Binary Rewards: A Comparative Study of Reward Design for Reinforcement Unlearningcited by 1
  14. 2026ST2U: Stateful Test-Time Unlearning via Restricted Knowledge Boundary Controlcited by 1
  15. 2025Hierarchical Dual-Strategy Unlearning for Biomedical and Healthcare Intelligence Using Imperfect and Privacy-Sensitive Medical Datacited by 0
  16. 2026textbfAGTAO: Robust and Stabilized LLM Unlearning via Adversarial Gating Training with Adaptive Orthogonalitycited by 0
  17. 2026Seamless Deception: Larger Language Models Are Better Knowledge Concealerscited by 0
  18. 2026oslash Source Models Leak What They Shouldn't nrightarrow: Unlearning Zero-Shot Transfer in Domain Adaptation Through Adversarial Optimizationcited by 0
  19. 2026WIN-U: Woodbury-Informed Newton-Unlearning as a retain-free Machine Unlearning Frameworkcited by 0
  20. 2026Gradient Concentration, Not Weight Saliency, Explains Representation-Level Class Unlearningcited by 0
  21. 2026Confess What You Know: Forget-Set Misalignment with Model Knowledge in LLM Unlearningcited by 0
  22. 2026RA2: Retain-Anchored Attraction Preserves Forget-Adjacent Utility in LLM Unlearningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Unlearning in LLMs: Methods, Evaluation, and Open Challenges
  2. 2024Towards Effective Evaluations and Comparisons for LLM Unlearning Methods
  3. 2025LLM Unlearning Under the Microscope: A Full-Stack View on Methods and Metrics
  4. 2025A Survey on Unlearning in Large Language Models
  5. 2024Rethinking Machine Unlearning for Large Language Models
  6. 2024Machine Unlearning in Large Language Models
  7. 2025SoK: Machine Unlearning for Large Language Models
  8. 2025OpenUnlearning: Accelerating LLM Unlearning via Unified Benchmarking of Methods and Metrics