mlsanityData to 7 Oct 2026

SOUL: Unlocking the Power of Second-Order Optimization for LLM Unlearning

arXiv 2404.18239 ↗First posted 28 Apr 2024cs.LG cs.CLcited by 104

On the map: LLM safety and security → Machine unlearning

Abstract (arXiv)

Large Language Models (LLMs) have highlighted the necessity of effective unlearning mechanisms to comply with data regulations and ethical AI practices. LLM unlearning aims at removing undesired data influences and associated model capabilities without compromising utility beyond the scope of unlearning. While interest in studying LLM unlearning is growing, the impact of the optimizer choice for LLM unlearning remains unexplored. In this work, we shed light on the significance of optimizer selection in LLM unlearning for the first time, establishing a clear connection between second-order optimization and influence unlearning (a classical approach using influence functions to update the model for data influence removal). This insight propels us to develop a second-order optimization-based LLM unlearning framework, termed Second-Order UnLearning (SOUL), which extends the static, one-shot model update using influence unlearning to a dynamic, iterative unlearning process. Our extensive experiments show that SOUL consistently outperforms conventional first-order methods across various unlearning tasks, models, and metrics, indicating that second-order optimization offers an effective and broadly applicable solution for LLM unlearning. Codes are available at https://github.com/OPTML-Group/SOUL.

Key facts. Cited by 104 papers in the map (data through 7 Oct 2026). Placed in Machine unlearning, one of 12 directions in LLM safety and security. Builds on 55 of its 79 references that are in the map. First posted on arXiv 28 Apr 2024.

Builds on

55 of 79 references.

Closest ancestors · 19

Same direction, newest first

  1. 2024Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning8 Apr
  2. 2024Challenging Forgets: Unveiling the Worst-Case Forget Sets in Machine Unlearning12 Mar
  3. 2024Guardrail Baselines for Unlearning in LLMs5 Mar
  4. 2024Inexact Unlearning Needs More Careful Evaluations to Avoid a False Sense of Privacy2 Mar
  5. 2024Eight Methods to Evaluate Robust Unlearning in LLMs26 Feb
  6. 2024Towards Safer Large Language Models through Machine Unlearning15 Feb
  7. 2024TOFU: A Task of Fictitious Unlearning for LLMs11 Jan
  8. 2023Unlearn What You Want to Forget: Efficient Unlearning for LLMs31 Oct
  9. 2023SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation19 Oct
  10. 2023In-Context Unlearning: Language Models as Few Shot Unlearners11 Oct
  11. 2023Who's Harry Potter? Approximate Unlearning in LLMs3 Oct
  12. 2023KGA: A General Machine Unlearning Framework Based on Knowledge Gap Alignment11 May
7 more
  1. 2023Model Sparsity Can Simplify Machine Unlearning11 Apr
  2. 2023Boundary Unlearning21 Mar
  3. 2022Knowledge Unlearning for Mitigating Privacy Risks in Language Models4 Oct
  4. 2021Unrolling SGD: Understanding Factors Influencing Machine Unlearning27 Sep
  5. 2021Machine Unlearning of Features and Labels26 Aug
  6. 2021Remember What You Want to Forget: Algorithms for Machine Unlearning4 Mar
  7. 2021Machine Unlearning via Algorithmic Stability25 Feb

Other references · 36

  1. 2023GPT-4 Technical Reportcited by 18,132
  2. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,491
  3. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  4. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,049
  5. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  6. 2019HellaSwag: Can a Machine Really Finish Your Sentence?cited by 4,018
  7. 2022OPT: Open Pre-trained Transformer Language Modelscited by 3,464
  8. 2016Pointer Sentinel Mixture Modelscited by 2,992
  9. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,919
  10. 2019PIQA: Reasoning about Physical Commonsense in Natural Languagecited by 2,818
  11. 2022Locating and Editing Factual Associations in GPTcited by 2,609
  12. 2023Sparks of Artificial General Intelligence: Early experiments with GPT-4cited by 2,579
24 more
  1. 2021Convex optimizationcited by 2,479
  2. 2019BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questionscited by 2,140
  3. 2018Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answeringcited by 2,114
  4. 2017Understanding Black-box Predictions via Influence Functionscited by 1,714
  5. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,235
  6. 2019Machine Unlearningcited by 1,020
  7. 2023BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Datasetcited by 828
  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. 2024The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearningcited by 488
  11. 2019Making AI Forget You: Data Deletion in Machine Learningcited by 447
  12. 2023Scalable Extraction of Training Data from (Production) Language Modelscited by 415
  13. 2023Studying Large Language Model Generalization with Influence Functionscited by 311
  14. 2020Amnesiac Machine Learningcited by 305
  15. 2023Can We Edit Factual Knowledge by In-Context Learning?cited by 284
  16. 2023Sophia: A Scalable Stochastic Second-order Optimizer for Language Model Pre-trainingcited by 260
  17. 2020Approximate Data Deletion from Machine Learning Modelscited by 250
  18. 2023Composing Parameter-Efficient Modules with Arithmetic Operationscited by 155
  19. 2019Privacy Risks of Securing Machine Learning Models against Adversarial Examplescited by 136
  20. 2016On Differentiating Parameterized Argmin and Argmax Problems with Application to Bi-level Optimizationcited by 108
  21. 2021Federated Unlearning via Class-Discriminative Pruningcited by 108
  22. 2023Identifying and Mitigating the Security Risks of Generative AIcited by 90
  23. 2024Machine Unlearning for Image-to-Image Generative Modelscited by 47
  24. 2020Federated Unlearningcited by 27

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 20; 2025: 50; 2026: 34.

104 papers.

Sort:

By area:

  1. 2024Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Modelscited by 164
  2. 2024Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearningcited by 160
  3. 2024Large Language Model Unlearning via Embedding-Corrupted Promptscited by 112
  4. 2024Reversing the Forget-Retain Objectives: An Efficient LLM Unlearning Framework from Logit Differencecited by 85
  5. 2024Unlearning or Obfuscating? Jogging the Memory of Unlearned LLMs via Benign Relearningcited by 81
  6. 2025OpenUnlearning: Accelerating LLM Unlearning via Unified Benchmarking of Methods and Metricscited by 80
  7. 2024Catastrophic Failure of LLM Unlearning via Quantizationcited by 76
  8. 2025Towards LLM Unlearning Resilient to Relearning Attacks: A Sharpness-Aware Minimization Perspective and Beyondcited by 76
  9. 2024LLM Unlearning via Loss Adjustment with Only Forget Datacited by 71
  10. 2024A Closer Look at Machine Unlearning for Large Language Modelscited by 54
  11. 2024On Large Language Model Continual Unlearningcited by 51
  12. 2024Meta-Unlearning on Diffusion Models: Preventing Relearning Unlearned Conceptscited by 39
38 more
  1. 2025Exploring Criteria of Loss Reweighting to Enhance LLM Unlearningcited by 39
  2. 2024Alternate Preference Optimization for Unlearning Factual Knowledge in Large Language Modelscited by 32
  3. 2024Benchmarking Vision Language Model Unlearning via Fictitious Facial Identity Datasetcited by 29
  4. 2024On Effects of Steering Latent Representation for Large Language Model Unlearningcited by 27
  5. 2024WAGLE: Strategic Weight Attribution for Effective and Modular Unlearning in Large Language Modelscited by 25
  6. 2024CLIPErase: Efficient Unlearning of Visual-Textual Associations in CLIPcited by 19
  7. 2025Learning to Unlearn while Retaining: Combating Gradient Conflicts in Machine Unlearningcited by 19
  8. 2025AgentA/B: Automated and Scalable Web A/BTesting with Interactive LLM Agentscited by 19
  9. 2025Shop-R1: Rewarding LLMs to Simulate Human Behavior in Online Shopping via Reinforcement Learningcited by 19
  10. 2025Tool Unlearning for Tool-Augmented LLMscited by 17
  11. 2025GUARD: Generation-time LLM Unlearning via Adaptive Restriction and Detectioncited by 17
  12. 2025Model Unlearning via Sparse Autoencoder Subspace Guided Projectionscited by 17
  13. 2025LLM Unlearning with LLM Beliefscited by 17
  14. 2024MU-Bench: A Multitask Multimodal Benchmark for Machine Unlearningcited by 16
  15. 2024MEOW: MEMOry Supervised LLM Unlearning Via Inverted Factscited by 16
  16. 2025Safety Mirage: How Spurious Correlations Undermine VLM Safety Fine-Tuning and Can Be Mitigated by Machine Unlearningcited by 16
  17. 2024Machine Unlearning Doesn't Do What You Think: Lessons for Generative AI Policy and Researchcited by 14
  18. 2025Reasoning Model Unlearning: Forgetting Traces, Not Just Answers, While Preserving Reasoning Skillscited by 12
  19. 2025Side Effects of Erasing Concepts from Diffusion Modelscited by 12
  20. 2025LLM Unlearning Reveals a Stronger-Than-Expected Coreset Effect in Current Benchmarkscited by 11
  21. 2025Distillation Robustifies Unlearningcited by 11
  22. 2025Unlearning Isn't Invisible: Detecting Unlearning Traces in LLMs from Model Outputscited by 10
  23. 2026Maximizing Local Entropy Where It Matters: Prefix-Aware Localized LLM Unlearningcited by 10
  24. 2025Agents Are All You Need for LLM Unlearningcited by 9
  25. 2025SoK: Machine Unlearning for Large Language Modelscited by 9
  26. 2025UCD: Unlearning in LLMs via Contrastive Decodingcited by 9
  27. 2026BalDRO: A Distributionally Robust Optimization based Framework for Large Language Model Unlearningcited by 8
  28. 2025Speech Unlearningcited by 7
  29. 2025DRAGON: Guard LLM Unlearning in Context via Negative Detection and Reasoningcited by 7
  30. 2025Not All Tokens Are Meant to Be Forgottencited by 6
  31. 2026Agentic Unlearning: When LLM Agent Meets Machine Unlearningcited by 6
  32. 2026Attention Smoothing Is All You Need For Unlearningcited by 6
  33. 2025Erasing Without Remembering: Implicit Knowledge Forgetting in Large Language Modelscited by 5
  34. 2025See, Think, Act: Online Shopper Behavior Simulation with VLM Agentscited by 5
  35. 2026EASE: Federated Multimodal Unlearning via Entanglement-Aware Anchor Closurecited by 5
  36. 2025SUV: Scalable Large Language Model Copyright Compliance with Regularized Selective Unlearningcited by 4
  37. 2025HessFormer: Hessians at Foundation Scalecited by 4
  38. 2025DUSK: Do Not Unlearn Shared Knowledgecited by 4

50 most cited of 104 All 104 (JSON)