mlsanityData to 14 Oct 2026

Evaluating Deep Unlearning in Large Language Models

arXiv 2410.15153 ↗First posted 19 Oct 2024cs.CLcited by 14

On the map: LLM safety and security → Machine unlearning

Abstract (arXiv)

Machine unlearning has emerged as an important component in developing safe and trustworthy models. Prior work on fact unlearning in LLMs has mostly focused on removing a specified target fact robustly, but often overlooks its deductive connections to other knowledge. We propose a new setting for fact unlearning, deep unlearning, where the goal is not only to remove a target fact but also to prevent it from being deduced via retained knowledge in the LLM and logical reasoning. We propose three novel metrics: Success-DU and Recall to measure unlearning efficacy, and Accuracy to measure the remainder model utility. To benchmark this setting, we leverage both (1) an existing real-world knowledge dataset, MQuAKE, that provides one-step deduction instances, and (2) newly construct a novel semi-synthetic dataset, Eval-DU, that allows multiple steps of realistic deductions among synthetic facts. Experiments reveal that current methods struggle with deep unlearning: they either fail to deeply unlearn, or excessively remove unrelated facts. Our results suggest that targeted algorithms may have to be developed for robust/deep fact unlearning in LLMs.

Key facts. Cited by 14 papers in the map (data through 14 Oct 2026). Placed in Machine unlearning, one of 12 directions in LLM safety and security. Builds on 45 of its 67 references that are in the map. First posted on arXiv 19 Oct 2024.

Builds on

45 of 67 references.

Closest ancestors · 13

Same direction, newest first

  1. 2024Breaking Chains: Unraveling the Links in Multi-Hop Knowledge Unlearning17 Oct
  2. 2024An Adversarial Perspective on Machine Unlearning for AI Safety26 Sep
  3. 2024MUSE: Machine Unlearning Six-Way Evaluation for Language Models8 Jul
  4. 2024Large Scale Knowledge Washing26 May
  5. 2024Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning8 Apr
  6. 2024Eight Methods to Evaluate Robust Unlearning in LLMs26 Feb
  7. 2024Machine Unlearning of Pre-trained Large Language Models23 Feb
  8. 2024Rethinking Machine Unlearning for Large Language Models13 Feb
  9. 2023Unlearn What You Want to Forget: Efficient Unlearning for LLMs31 Oct
  10. 2023Large Language Model Unlearning14 Oct
  11. 2023In-Context Unlearning: Language Models as Few Shot Unlearners11 Oct
  12. 2023Who's Harry Potter? Approximate Unlearning in LLMs3 Oct
1 more
  1. 2022Knowledge Unlearning for Mitigating Privacy Risks in Language Models4 Oct

Other references · 32

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  3. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  4. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  5. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  6. 2019PIQA: Reasoning about Physical Commonsense in Natural Languagecited by 2,883
  7. 2022Locating and Editing Factual Associations in GPTcited by 2,691
  8. 2019HuggingFace's Transformers: State-of-the-art Natural Language Processingcited by 2,218
  9. 2020Extracting Training Data from Large Language Modelscited by 2,189
  10. 2022Editing Models with Task Arithmeticcited by 1,120
  11. 2022Mass-Editing Memory in a Transformercited by 1,001
  12. 2014Embedding Entities and Relations for Learning and Inference in Knowledge Basescited by 932
20 more
  1. 2017RACE: Large-scale ReAding Comprehension Dataset From Examinationscited by 861
  2. 2020Knowledge Graphscited by 621
  3. 2023Textbooks Are All You Need II: phi-1.5 technical reportcited by 565
  4. 2021Editing Factual Knowledge in Language Modelscited by 553
  5. 2019Certified Data Removal from Machine Learning Modelscited by 535
  6. 2024The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearningcited by 509
  7. 2019Making AI Forget You: Data Deletion in Machine Learningcited by 465
  8. 2023Scalable Extraction of Training Data from (Production) Language Modelscited by 434
  9. 2019Neural Probabilistic Logic Programming in DeepProbLogcited by 394
  10. 2023StructGPT: A General Framework for Large Language Model to Reason over Structured Datacited by 343
  11. 2023MQuAKE: Assessing Knowledge Editing in Language Models via Multi-Hop Questionscited by 300
  12. 2015A Review of Relational Machine Learning for Knowledge Graphscited by 265
  13. 2017End-to-End Differentiable Provingcited by 161
  14. 2023Can Sensitive Information Be Deleted From LLMs? Objectives for Defending Against Extraction Attackscited by 158
  15. 2023Knowledge-Augmented Language Model Prompting for Zero-Shot Knowledge Graph Question Answeringcited by 119
  16. 2019Probabilistic Logic Neural Networks for Reasoningcited by 69
  17. 2016TensorLog: A Differentiable Deductive Databasecited by 29
  18. 2023ChatRule: Mining Logical Rules with Large Language Models for Knowledge Graph Reasoningcited by 26
  19. 2023Neural Compositional Rule Learning for Knowledge Graph Reasoningcited by 20
  20. 2022Ruleformer: Context-aware Differentiable Rule Mining over Knowledge Graphcited by 9

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 11; 2026: 3.

14 papers.

Sort:

By area:

  1. 2025Do LLMs Really Forget? Evaluating Unlearning with Knowledge Correlation and Confidence Awarenesscited by 14
  2. 2025ForgetMe: Evaluating Selective Forgetting in Generative Modelscited by 11
  3. 2025Agents Are All You Need for LLM Unlearningcited by 10
  4. 2025Keeping an Eye on LLM Unlearning: The Hidden Risk and Remedycited by 9
  5. 2025LUNE: Efficient LLM Unlearning via LoRA Fine-Tuning with Negative Examplescited by 8
  6. 2025The Trojan Knowledge: Bypassing Commercial LLM Guardrails via Harmless Prompt Weaving and Adaptive Tree Searchcited by 7
  7. 2025On the Impossibility of Retrain Equivalence in Machine Unlearningcited by 6
  8. 2025Erasing Without Remembering: Implicit Knowledge Forgetting in Large Language Modelscited by 5
  9. 2025ThinkEval: Practical Evaluation of Knowledge Leakage in LLM Editing using Thought-based Knowledge Graphscited by 4
  10. 2025Learning-Time Encoding Shapes Unlearning in LLMscited by 1
  11. 2025WaterDrum: Watermarking for Data-centric Unlearning Metriccited by 0
  12. 2026Behemoth: Benchmarking Unlearning in LLMs Using Fully Synthetic Datacited by 0
2 more
  1. 2026What to Forget in Unlearning? Forget Set Curation for Language Modelscited by 0
  2. 2026Bounding Retraining Equivalence and the Deletion Floor in Materials Machine Unlearningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Making LLMs Truly Forget: Deep Unlearning by Searching, Selecting, and Severing Knowledge Paths
  2. 2026Unlearning in LLMs: Methods, Evaluation, and Open Challenges
  3. 2025A Survey on Unlearning in Large Language Models
  4. 2024Towards Effective Evaluations and Comparisons for LLM Unlearning Methods
  5. 2026Measuring the Depth of LLM Unlearning via Activation Patching
  6. 2024Catastrophic Failure of LLM Unlearning via Quantization
  7. 2026Unlearners Can Lie: Evaluating and Improving Honesty in LLM Unlearning
  8. 2025Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs