mlsanityData to 14 Oct 2026

SUGAR: A Sweeter Spot for Generative Unlearning of Many Identities

arXiv 2512.06562 ↗First posted 6 Dec 2025cs.CV cs.AIcited by 1

On the map: LLM safety and security → Vision-language and text-to-image safety

Abstract (arXiv)

Recent advances in 3D-aware generative models have enabled high-fidelity image synthesis of human identities. However, this progress raises urgent questions around user consent and the ability to remove specific individuals from a model's output space. We address this by introducing SUGAR, a framework for scalable generative unlearning that enables the removal of many identities (simultaneously or sequentially) without retraining the entire model. Rather than projecting unwanted identities to unrealistic outputs or relying on static template faces, SUGAR learns a personalized surrogate latent for each identity, diverting reconstructions to visually coherent alternatives while preserving the model's quality and diversity. We further introduce a continual utility preservation objective that guards against degradation as more identities are forgotten. SUGAR achieves state-of-the-art performance in removing up to 200 identities, while delivering up to a 700% improvement in retention utility compared to existing baselines. Our code is publicly available at https://github.com/judydnguyen/SUGAR-Generative-Unlearn.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Vision-language and text-to-image safety, one of 12 directions in LLM safety and security. Builds on 39 of its 51 references that are in the map. First posted on arXiv 6 Dec 2025.

Builds on

39 of 51 references.

Closest ancestors · 3

Same direction, newest first

  1. 2024Generative Unlearning for Any Identity16 May
  2. 2024UnlearnCanvas: Stylized Image Dataset for Enhanced Machine Unlearning Evaluation in Diffusion Models19 Feb
  3. 2023Adapt then Unlearn: Exploring Parameter Space Semantics for Unlearning in Generative Adversarial Networks25 Sep

Other references · 36

  1. 2017Attention Is All You Needcited by 60,361
  2. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  3. 2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 6,013
  4. 2016Overcoming catastrophic forgetting in neural networkscited by 4,787
  5. 2017Progressive Growing of GANs for Improved Quality, Stability, and Variationcited by 2,988
  6. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,281
  7. 2021Efficient Geometry-aware 3D Generative Adversarial Networkscited by 1,205
  8. 2019Machine Unlearningcited by 1,061
  9. 2021StyleCLIP: Text-Driven Manipulation of StyleGAN Imagerycited by 991
  10. 2023Extracting Training Data from Diffusion Modelscited by 732
  11. 2019Eternal Sunshine of the Spotless Net: Selective Forgetting in Deep Networkscited by 618
  12. 2023Erasing Concepts from Diffusion Modelscited by 539
24 more
  1. 2022Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Modelscited by 424
  2. 2020InterFaceGAN: Interpreting the Disentangled Face Representation Learned by GANscited by 394
  3. 2020CurricularFace: Adaptive Curriculum Learning Loss for Deep Face Recognitioncited by 321
  4. 2021Membership Inference Attacks on Machine Learning: A Surveycited by 321
  5. 2024MACE: Mass Concept Erasure in Diffusion Modelscited by 307
  6. 2023Large Language Model Unlearningcited by 302
  7. 2021Fast Yet Effective Machine Unlearningcited by 217
  8. 2020Forgetting Outside the Box: Scrubbing Deep Networks of Information Accessible from Input-Output Observationscited by 180
  9. 2023Selective Amnesia: A Continual Learning Approach to Forgetting in Deep Generative Modelscited by 177
  10. 2021Adaptive Machine Unlearningcited by 176
  11. 2019Photorealistic Style Transfer via Wavelet Transformscited by 166
  12. 2020Mixed-Privacy Forgetting in Deep Networkscited by 163
  13. 2021StyleSwin: Transformer-based GAN for High-resolution Image Generationcited by 152
  14. 2022Zero-Shot Machine Unlearningcited by 133
  15. 2023Machine Unlearning: Solutions and Challengescited by 93
  16. 2023Right to be Forgotten in the Era of Large Language Models: Implications, Challenges, and Solutionscited by 60
  17. 2022Generative Multiplane Images: Making a 2D GAN 3D-Awarecited by 57
  18. 2021Text to Image Generation with Semantic-Spatial Aware GANcited by 53
  19. 2024Single Image Unlearning: Efficient Machine Unlearning in Multimodal Large Language Modelscited by 46
  20. 2023MetaCloak: Preventing Unauthorized Subject-driven Text-to-image Diffusion-based Synthesis via Meta-learningcited by 44
  21. 2023GenAI Against Humanity: Nefarious Applications of Generative Artificial Intelligence and Large Language Modelscited by 42
  22. 2023Make Encoder Great Again in 3D GAN Inversion through Geometry and Occlusion-Aware Encodingcited by 35
  23. 2024A Review on Machine Unlearningcited by 32
  24. 2024Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Modelscited by 12

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Damnatio Memoriae: Adversarially and Selectively Forgetting Identities in the Embedding Space of Face Recognition Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026PIU: Proximity-guided Identity Unlearning in ID-Conditioned Diffusion Models
  2. 2025Beyond the Pixels: VLM-based Evaluation of Identity Preservation in Reference-Guided Synthesis
  3. 2025UNO: Unlearning via Orthogonalization in Generative models
  4. 2024CapHuman: Capture Your Moments in Parallel Universes
  5. 2026Unlearning the Unpromptable: Prompt-free Instance Unlearning in Diffusion Models
  6. 2025Reverse Personalization
  7. 2024Fusion is all you need: Face Fusion for Customized Identity-Preserving Image Synthesis
  8. 2024InstantID: Zero-shot Identity-Preserving Generation in Seconds