mlsanityData to 14 Oct 2026

Towards Safer Large Language Models through Machine Unlearning

arXiv 2402.10058 ↗First posted 15 Feb 2024cs.CLcited by 139

On the map: LLM safety and security → Machine unlearning

Abstract (arXiv)

The rapid advancement of Large Language Models (LLMs) has demonstrated their vast potential across various domains, attributed to their extensive pretraining knowledge and exceptional generalizability. However, LLMs often encounter challenges in generating harmful content when faced with problematic prompts. To address this problem, existing work attempted to implement a gradient ascent based approach to prevent LLMs from producing harmful output. While these methods can be effective, they frequently impact the model utility in responding to normal prompts. To address this gap, we introduce Selective Knowledge negation Unlearning (SKU), a novel unlearning framework for LLMs, designed to eliminate harmful knowledge while preserving utility on normal prompts. Specifically, SKU is consisted of two stages: harmful knowledge acquisition stage and knowledge negation stage. The first stage aims to identify and acquire harmful knowledge within the model, whereas the second is dedicated to remove this knowledge. SKU selectively isolates and removes harmful knowledge in model parameters, ensuring the model's performance remains robust on normal prompts. Our experiments conducted across various LLM architectures demonstrate that SKU identifies a good balance point between removing harmful information and preserving utility.

Key facts. Cited by 139 papers in the map (data through 14 Oct 2026). Placed in Machine unlearning, one of 12 directions in LLM safety and security. Builds on 32 of its 42 references that are in the map. First posted on arXiv 15 Feb 2024.

Builds on

32 of 42 references.

Closest ancestors · 7

Same direction, newest first

  1. 2024TOFU: A Task of Fictitious Unlearning for LLMs11 Jan
  2. 2023Unlearn What You Want to Forget: Efficient Unlearning for LLMs31 Oct
  3. 2023Breaking the Trilemma of Privacy, Utility, Efficiency via Controllable Machine Unlearning28 Oct
  4. 2023Who's Harry Potter? Approximate Unlearning in LLMs3 Oct
  5. 2022Can Bad Teaching Induce Forgetting? Unlearning in Deep Networks using an Incompetent Teacher17 May
  6. 2021Unrolling SGD: Understanding Factors Influencing Machine Unlearning27 Sep
  7. 2021Remember What You Want to Forget: Algorithms for Machine Unlearning4 Mar

Other references · 25

  1. 2020Language Models are Few-Shot Learnerscited by 32,419
  2. 2023GPT-4 Technical Reportcited by 18,396
  3. 2022Training language models to follow instructions with human feedbackcited by 15,971
  4. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  5. 2022PaLM: Scaling Language Modeling with Pathwayscited by 5,326
  6. 2022Large Language Models are Zero-Shot Reasonerscited by 4,374
  7. 2022OPT: Open Pre-trained Transformer Language Modelscited by 3,512
  8. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,975
  9. 2023Code Llama: Open Foundation Models for Codecited by 2,139
  10. 2022Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?cited by 1,434
  11. 2022Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference timecited by 1,396
  12. 2020BLEURT: Learning Robust Metrics for Text Generationcited by 1,203
13 more
  1. 2019Machine Unlearningcited by 1,061
  2. 2023BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Datasetcited by 860
  3. 2021Robust fine-tuning of zero-shot modelscited by 797
  4. 2023Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedbackcited by 759
  5. 2019Linear Mode Connectivity and the Lottery Ticket Hypothesiscited by 637
  6. 2021Merging Models with Fisher-Weighted Averagingcited by 621
  7. 2019Certified Data Removal from Machine Learning Modelscited by 535
  8. 2023Is ChatGPT a General-Purpose Natural Language Processing Task Solver?cited by 488
  9. 2022Git Re-Basin: Merging Models modulo Permutation Symmetriescited by 471
  10. 2019Making AI Forget You: Data Deletion in Machine Learningcited by 465
  11. 2021Do Prompt-Based Models Really Understand the Meaning of their Prompts?cited by 326
  12. 2022Patching open-vocabulary models by interpolating weightscited by 190
  13. 2024Democratizing Large Language Models via Personalized Parameter-Efficient Fine-tuningcited by 90

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 50; 2025: 61; 2026: 28.

139 papers.

Sort:

By area:

  1. 2024The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearningcited by 509
  2. 2024Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearningcited by 504
  3. 2024Model Merging in LLMs, MLLMs, and Beyond: Methods, Theories, Applications and Opportunitiescited by 254
  4. 2024Latent Adversarial Training Improves Robustness to Persistent Harmful Behaviors in LLMscited by 173
  5. 2024Eight Methods to Evaluate Robust Unlearning in LLMscited by 163
  6. 2024An Adversarial Perspective on Machine Unlearning for AI Safetycited by 123
  7. 2024Large Language Model Unlearning via Embedding-Corrupted Promptscited by 116
  8. 2024SOUL: Unlocking the Power of Second-Order Optimization for LLM Unlearningcited by 104
  9. 2024Catastrophic Failure of LLM Unlearning via Quantizationcited by 81
  10. 2025Towards LLM Unlearning Resilient to Relearning Attacks: A Sharpness-Aware Minimization Perspective and Beyondcited by 79
  11. 2024Personalized Pieces: Efficient Personalized Large Language Models through Collaborative Effortscited by 74
  12. 2026Memory for Autonomous LLM Agents:Mechanisms, Evaluation, and Emerging Frontierscited by 65
38 more
  1. 2024Protecting Privacy in Multimodal Large Language Models with MLLMU-Benchcited by 64
  2. 2024A Closer Look at Machine Unlearning for Large Language Modelscited by 56
  3. 2024UnUnlearning: Unlearning is not sufficient for content regulation in advanced generative AIcited by 51
  4. 2024Machine Unlearning in Generative AI: A Surveycited by 49
  5. 2024Meta-Unlearning on Diffusion Models: Preventing Relearning Unlearned Conceptscited by 40
  6. 2024Threats, Attacks, and Defenses in Machine Unlearning: A Surveycited by 39
  7. 2024CLEAR: Character Unlearning in Textual and Visual Modalitiescited by 39
  8. 2025AdaSteer: Your Aligned LLM is Inherently an Adaptive Jailbreak Defendercited by 38
  9. 2024Low-Rank Adaptation for Foundation Models: A Comprehensive Reviewcited by 37
  10. 2024Revisiting Who's Harry Potter: Towards Targeted Unlearning from a Causal Intervention Perspectivecited by 36
  11. 2024Towards Robust and Parameter-Efficient Knowledge Unlearning for LLMscited by 36
  12. 2024Machine Unlearning: A Comprehensive Surveycited by 35
  13. 2025A Comprehensive Survey of Machine Unlearning Techniques for Large Language Modelscited by 34
  14. 2025Towards Trustworthy GUI Agents: A Surveycited by 34
  15. 2024Alternate Preference Optimization for Unlearning Factual Knowledge in Large Language Modelscited by 33
  16. 2025SafeEraser: Enhancing Safety in Multimodal Large Language Models through Multimodal Machine Unlearningcited by 32
  17. 2025Towards Trustworthy Retrieval Augmented Generation for Large Language Models: A Surveycited by 30
  18. 2025MMUnlearner: Reformulating Multimodal Machine Unlearning in the Era of Multimodal Large Language Modelscited by 30
  19. 2024Intrinsic Test of Unlearning Using Parametric Knowledge Tracescited by 29
  20. 2024From Theft to Bomb-Making: The Ripple Effect of Unlearning in Defending Against Jailbreak Attackscited by 29
  21. 2024To Forget or Not? Towards Practical Knowledge Unlearning for Large Language Modelscited by 26
  22. 2024WAGLE: Strategic Weight Attribution for Effective and Modular Unlearning in Large Language Modelscited by 26
  23. 2025Modality-Aware Neuron Pruning for Unlearning in Multimodal Large Language Modelscited by 26
  24. 2024Unified Parameter-Efficient Unlearning for LLMscited by 25
  25. 2024Avoiding Copyright Infringement via Large Language Model Unlearningcited by 24
  26. 2024When Machine Unlearning Meets Retrieval-Augmented Generation (RAG): Keep Secret or Forget Knowledge?cited by 24
  27. 2024Model Surgery: Modulating LLM's Behavior Via Simple Parameter Editingcited by 23
  28. 2025BLUR: A Bi-Level Optimization Approach for LLM Unlearningcited by 23
  29. 2025ReLearn: Unlearning via Learning for Large Language Modelscited by 22
  30. 2024CLIPErase: Efficient Unlearning of Visual-Textual Associations in CLIPcited by 19
  31. 2024Split, Unlearn, Merge: Leveraging Data Attributes for More Effective Unlearning in LLMscited by 18
  32. 2024Large Action Models: From Inception to Implementationcited by 18
  33. 2024Towards Comprehensive Post Safety Alignment of Large Language Models via Safety Patchingcited by 17
  34. 2024Towards Efficient and Effective Unlearning of Large Language Models for Recommendationcited by 16
  35. 2024Large Scale Knowledge Washingcited by 16
  36. 2024UnStar: Unlearning with Self-Taught Anti-Sample Reasoning for LLMscited by 15
  37. 2024Machine Unlearning Doesn't Do What You Think: Lessons for Generative AI Policy and Researchcited by 15
  38. 2024Unveiling Entity-Level Unlearning for Large Language Models: A Comprehensive Analysiscited by 14

50 most cited of 139 All 139 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2024Machine Unlearning in Large Language Models
  2. 2025Not Every Token Needs Forgetting: Selective Unlearning to Limit Change in Utility in Large Language Model Unlearning
  3. 2024Machine Unlearning in Large Language Models
  4. 2024Soft Prompting for Unlearning in Large Language Models
  5. 2025LLM Unlearning using Gradient Ratio-Based Influence Estimation and Noise Injection
  6. 2024Machine Unlearning of Pre-trained Large Language Models
  7. 2023Knowledge Unlearning for LLMs: Tasks, Methods, and Challenges
  8. 2024Multi-Objective Large Language Model Unlearning