mlsanityData to 7 Oct 2026

Memory Self-Regeneration: Uncovering Hidden Knowledge in Unlearned Models

arXiv 2510.03263 ↗First posted 26 Sep 2025cs.LG cs.AIcited by 2

On the map: LLM safety and security → Machine unlearning

Abstract (arXiv)

The impressive capability of modern text-to-image models to generate realistic visuals has come with a serious drawback: they can be misused to create harmful, deceptive or unlawful content. This has accelerated the push for machine unlearning. This new field seeks to selectively remove specific knowledge from a model's training data without causing a drop in its overall performance. However, it turns out that actually forgetting a given concept is an extremely difficult task. Models exposed to attacks using adversarial prompts show the ability to generate so-called unlearned concepts, which can be not only harmful but also illegal. In this paper, we present considerations regarding the ability of models to forget and recall knowledge, introducing the Memory Self-Regeneration task. Furthermore, we present MemoRa strategy, which we consider to be a regenerative approach supporting the effective recovery of previously lost knowledge. Moreover, we propose that robustness in knowledge retrieval is a crucial yet underexplored evaluation measure for developing more robust and effective unlearning techniques. Finally, we demonstrate that forgetting occurs in two distinct ways: short-term, where concepts can be quickly recalled, and long-term, where recovery is more challenging. Code is available at https://gmum.github.io/MemoRa/.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Machine unlearning, one of 12 directions in LLM safety and security. Builds on 35 of its 51 references that are in the map. First posted on arXiv 26 Sep 2025.

Builds on

35 of 51 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025SEMU: Singular Value Decomposition for Efficient Machine Unlearning11 Feb
  2. 2024Unlearning or Obfuscating? Jogging the Memory of Unlearned LLMs via Benign Relearning19 Jun
  3. 2024Scissorhands: Scrub Data Influence via Connection Sensitivity in Networks11 Jan
  4. 2023SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation19 Oct
  5. 2023Towards Unbounded Machine Unlearning20 Feb

Other references · 30

  1. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  2. 2022Classifier-Free Diffusion Guidancecited by 5,717
  3. 2023Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detectioncited by 3,274
  4. 2021CLIPScore: A Reference-free Evaluation Metric for Image Captioningcited by 2,388
  5. 2022An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversioncited by 2,231
  6. 2015Microsoft COCO Captions: Data Collection and Evaluation Servercited by 1,447
  7. 2023Erasing Concepts from Diffusion Modelscited by 530
  8. 2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Modelscited by 501
  9. 2023Unified Concept Editing in Diffusion Modelscited by 377
  10. 2018TextBugger: Generating Adversarial Text Against Real-world Applicationscited by 376
  11. 2020BAE: BERT-based Adversarial Examples for Text Classificationcited by 316
  12. 2023Ablating Concepts in Text-to-Image Diffusion Modelscited by 312
18 more
  1. 2024MACE: Mass Concept Erasure in Diffusion Modelscited by 298
  2. 2023Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Modelscited by 297
  3. 2022Red-Teaming the Stable Diffusion Safety Filtercited by 257
  4. 2023To Generate or Not? Safety-Driven Unlearned Diffusion Models Are Still Easy To Generate Unsafe Images ... For Nowcited by 204
  5. 2023Selective Amnesia: A Continual Learning Approach to Forgetting in Deep Generative Modelscited by 175
  6. 2023One-Dimensional Adapter to Rule Them All: Concepts, Diffusion Models and Erasing Applicationscited by 164
  7. 2024Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Modelscited by 164
  8. 2024Semantic Image Inversion and Editing using Rectified Stochastic Differential Equationscited by 162
  9. 2023Prompting4Debugging: Red-Teaming Text-to-Image Diffusion Models by Finding Problematic Promptscited by 161
  10. 2022The Privacy Onion Effect: Memorization is Relativecited by 128
  11. 2025SAeUron: Interpretable Concept Unlearning in Diffusion Models with Sparse Autoencoderscited by 75
  12. 2020From Hero to Z'eroe: A Benchmark of Low-Level Adversarial Attackscited by 36
  13. 2025Classifier-free Guidance with Adaptive Scalingcited by 19
  14. 2019Using Small Proxy Datasets to Accelerate Hyperparameter Searchcited by 13
  15. 2025Erased but Not Forgotten: How Backdoors Compromise Concept Erasurecited by 9
  16. 2024AutoLoRA: AutoGuidance Meets Low-Rank Adaptation for Diffusion Modelscited by 7
  17. 2025UnGuide: Learning to Forget with LoRA-Guided Diffusion Modelscited by 7
  18. 2025Automating Evaluation of Diffusion Model Unlearning with (Vision-) Language Model World Knowledgecited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026ReLAPSe: Reinforcement-Learning-trained Adversarial Prompt Search for Erased concepts in unlearned diffusion modelscited by 0
  2. 2026Concept Score Relearning: A Unified Cross-Architecture Attack on Concept Erasurecited by 0