mlsanityData to 14 Oct 2026

Data Unlearning in Diffusion Models

arXiv 2503.01034 ↗First posted 2 Mar 2025cs.LGcited by 18

On the map: Diffusion models → Robustness, privacy and adversarial aspects

also in 1 more
Abstract (arXiv)

Recent work has shown that diffusion models memorize and reproduce training data examples. At the same time, large copyright lawsuits and legislation such as GDPR have highlighted the need for erasing datapoints from diffusion models. However, retraining from scratch is often too expensive. This motivates the setting of data unlearning, i.e., the study of efficient techniques for unlearning specific datapoints from the training set. Existing concept unlearning techniques require an anchor prompt/class/distribution to guide unlearning, which is not available in the data unlearning setting. General-purpose machine unlearning techniques were found to be either unstable or failed to unlearn data. We therefore propose a family of new loss functions called Subtracted Importance Sampled Scores (SISS) that utilize importance sampling and are the first method to unlearn data with theoretical guarantees. SISS is constructed as a weighted combination between simpler objectives that are responsible for preserving model quality and unlearning the targeted datapoints. When evaluated on CelebA-HQ and MNIST, SISS achieved Pareto optimality along the quality and unlearning strength dimensions. On Stable Diffusion, SISS successfully mitigated memorization on nearly 90% of the prompts we tested.

Key facts. Cited by 18 papers in the map (data through 14 Oct 2026). Placed in Robustness, privacy and adversarial aspects, one of 14 directions in Diffusion models. Builds on 34 of its 42 references that are in the map. First posted on arXiv 2 Mar 2025.

Builds on

34 of 42 references.

Closest ancestors · 9

Same direction, newest first

  1. 2023Privacy Threats in Stable Diffusion Models15 Nov
  2. 2023SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation19 Oct
  3. 2023Unified Concept Editing in Diffusion Models25 Aug
  4. 2023A Reproducible Extraction of Training Images from Diffusion Models15 May
  5. 2023Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Models30 Mar
  6. 2023Ablating Concepts in Text-to-Image Diffusion Models23 Mar
  7. 2023Erasing Concepts from Diffusion Models13 Mar
  8. 2023Extracting Training Data from Diffusion Models30 Jan
  9. 2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Models9 Nov

Other references · 25

  1. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  3. 2020Denoising Diffusion Implicit Modelscited by 9,781
  4. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 9,122
  5. 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,349
  6. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,152
  7. 2017Progressive Growing of GANs for Improved Quality, Stability, and Variationcited by 2,988
  8. 2016Improved Techniques for Training GANscited by 2,970
  9. 2019Machine Unlearningcited by 1,061
  10. 2022Exploring CLIP for Assessing the Look and Feel of Imagescited by 991
  11. 2019Eternal Sunshine of the Spotless Net: Selective Forgetting in Deep Networkscited by 618
  12. 2019Making AI Forget You: Data Deletion in Machine Learningcited by 465
13 more
  1. 2022Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Modelscited by 424
  2. 2020Approximate Data Deletion from Machine Learning Modelscited by 255
  3. 2023Understanding and Mitigating Copying in Diffusion Modelscited by 209
  4. 2023Selective Amnesia: A Continual Learning Approach to Forgetting in Deep Generative Modelscited by 177
  5. 2022A Self-Supervised Descriptor for Image Copy Detectioncited by 176
  6. 2024Detecting, Explaining, and Mitigating Memorization in Diffusion Modelscited by 95
  7. 2023Exploring the Landscape of Machine Unlearning: A Comprehensive Survey and Taxonomycited by 51
  8. 2024Machine Unlearning for Image-to-Image Generative Modelscited by 47
  9. 2022Deep Regression Unlearningcited by 32
  10. 2023Training Data Protection with Compositional Diffusion Modelscited by 17
  11. 2024Who's in and who's out? A case study of multimodal CLIP-filtering in DataCompcited by 15
  12. 2023Data Redaction from Conditional Generative Modelscited by 8
  13. 2022Approximate Data Deletion in Generative Modelscited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 8; 2026: 10.

18 papers.

Sort:

By area:

  1. 2025wd1: Weighted Policy Optimization for Reasoning in Diffusion Language Modelscited by 71
  2. 2025ResearchCodeBench: Benchmarking LLMs on Implementing Novel Machine Learning Research Codecited by 36
  3. 2025Alignment and Safety in Large Language Models: Safety Mechanisms, Training Paradigms, and Emerging Challengescited by 15
  4. 2025Erasing Concepts, Steering Generations: A Comprehensive Survey of Concept Suppressioncited by 10
  5. 2026Is your algorithm unlearning or untraining?cited by 4
  6. 2026Active Flow Expansion for Out-of-Distribution Discovery: from Theory to Moleculescited by 3
  7. 2025Not Every Time and Frequency Need to Be Forgotten in Diffusion Unlearningcited by 2
  8. 2026Unlearning in Diffusion Models: A Unified Framework with KL Divergence and Likelihood Constraintscited by 2
  9. 2025GIFT: Gradient-aware Immunization of diffusion models against malicious Fine-Tuning with safe concepts retentioncited by 1
  10. 2025IMU: Influence-guided Machine Unlearningcited by 1
  11. 2026Unlearning the Unpromptable: Prompt-free Instance Unlearning in Diffusion Modelscited by 1
  12. 2026Training data attribution in diffusion models via mirrored unlearning and noise-consistent skewcited by 1
6 more
  1. 2026PIU: Proximity-guided Identity Unlearning in ID-Conditioned Diffusion Modelscited by 1
  2. 2025Adaptive-lambda Subtracted Importance Sampled Scores in Machine Unlearning for DDPMs and VAEscited by 0
  3. 2026You Don't Need All That Attention: Surgical Memorization Mitigation in Text-to-Image Diffusion Modelscited by 0
  4. 2026Mitigating Sequential Reappearance in Diffusion Data-Point Unlearningcited by 0
  5. 2026Continual Data Unlearning in Diffusion Models via Transition-based Regularizationcited by 0
  6. 2026Data Unlearning via Inverse Distillationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Model Integrity when Unlearning with T2I Diffusion Models
  2. 2025ReTrack: Data Unlearning in Diffusion Models through Redirecting the Denoising Trajectory
  3. 2024Remaining-data-free Machine Unlearning by Suppressing Sample Contribution
  4. 2026Forgetting Similar Samples: Can Machine Unlearning Do it Better?
  5. 2025Certified Data Removal Under High-dimensional Settings
  6. 2025Efficient Machine Unlearning by Model Splitting and Core Sample Selection
  7. 2026Efficient machine unlearning with minimax optimality
  8. 2025Machine Unlearning under Overparameterization