mlsanityData to 7 Oct 2026

Understanding and Mitigating Memorization in Generative Models via Sharpness of Probability Landscapes

arXiv 2412.04140 ↗First posted 5 Dec 2024cs.LG cs.AIcited by 20

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

In this paper, we introduce a geometric framework to analyze memorization in diffusion models through the sharpness of the log probability density. We mathematically justify a previously proposed score-difference-based memorization metric by demonstrating its effectiveness in quantifying sharpness. Additionally, we propose a novel memorization metric that captures sharpness at the initial stage of image generation in latent diffusion models, offering early insights into potential memorization. Leveraging this metric, we develop a mitigation strategy that optimizes the initial noise of the generation process using a sharpness-aware regularization term. The code is publicly available at https://github.com/Dongjae0324/sharpness_memorization_diffusion.

Key facts. Cited by 20 papers in the map (data through 7 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 36 of its 49 references that are in the map. First posted on arXiv 5 Dec 2024.

Builds on

36 of 49 references.

Closest ancestors · 18

Same direction, newest first

  1. 2024Exploring Local Memorization in Diffusion Models via Bright Ending Attention29 Oct
  2. 2024Manifolds, Random Matrices and Spectral Gaps: The geometric phases of generative diffusion8 Oct
  3. 2024Detecting, Explaining, and Mitigating Memorization in Diffusion Models31 Jul
  4. 2024A Geometric View of Data Complexity: Efficient Local Intrinsic Dimension Estimation with Diffusion Models5 Jun
  5. 2024Unveiling and Mitigating Memorization in Text-to-image Diffusion Models through Cross Attention17 Mar
  6. 2024On gauge freedom, conservativity and intrinsic dimensionality estimation in diffusion models6 Feb
  7. 2023Score-based generative models learn manifold-like structures with constrained mixing16 Nov
  8. 2023On Memorization in Diffusion Models4 Oct
  9. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis4 Jul
  10. 2023Understanding and Mitigating Copying in Diffusion Models31 May
  11. 2022Classifier-Free Diffusion Guidance26 Jul
  12. 2022Maximum Likelihood Training for Score-Based Diffusion ODEs by High-Order Denoising Score Matching16 Jun
6 more
  1. 2022Score-Based Generative Models Detect Manifolds2 Jun
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
  3. 2021Maximum Likelihood Training of Score-Based Diffusion Models22 Jan
  4. 2020Score-Based Generative Modeling through Stochastic Differential Equations26 Nov
  5. 2020Denoising Diffusion Implicit Models6 Oct
  6. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 18

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  3. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 13,958
  4. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  5. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,893
  6. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,322
  7. 2022Video Diffusion Modelscited by 2,280
  8. 2023Extracting Training Data from Diffusion Modelscited by 723
  9. 2022Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Modelscited by 418
  10. 2021The Intrinsic Dimension of Images and Its Impact on Learningcited by 307
  11. 2022A Self-Supervised Descriptor for Image Copy Detectioncited by 175
  12. 2022Generating Long Videos of Dynamic Scenescited by 113
6 more
  1. 2021Estimating High Order Gradients of the Data Distribution by Denoisingcited by 63
  2. 2023A Reproducible Extraction of Training Images from Diffusion Modelscited by 60
  3. 2024A Geometric Framework for Understanding Memorization in Generative Modelscited by 40
  4. 2022LIDL: Local Intrinsic Dimension Estimation Using Approximate Likelihoodcited by 30
  5. 2024Diffusion Models in textitDe Novo Drug Designcited by 26
  6. 2023Data-Copying in Generative Models: A Formal Frameworkcited by 9

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 6; 2026: 14.

20 papers.

Sort:

By area:

  1. 2026Making Reconstruction FID Predictive of Diffusion Generation FIDcited by 12
  2. 2025Generalization of Diffusion Models Arises with a Balanced Representation Spacecited by 10
  3. 2025How Diffusion Models Memorizecited by 6
  4. 2025Not Every Time and Frequency Need to Be Forgotten in Diffusion Unlearningcited by 2
  5. 2025Transport Novelty Distance: A Distributional Metric for Evaluating Material Generative Modelscited by 2
  6. 2026Detecting and Mitigating Memorization in Diffusion Models through Anisotropy of the Log-Probabilitycited by 2
  7. 2026Substitution-Based Analysis of Structural Novelty for Generative Models of Materialscited by 2
  8. 2025Dominant vs. Dominated: Concept-Level Generative Collapse in Diffusion Modelscited by 1
  9. 2026Local Hessian Spectral Filtering for Robust Intrinsic Dimension Estimationcited by 1
  10. 2025Identifying Memorization of Diffusion Models through p-Laplace Analysis: Estimators, Bounds and Applicationscited by 0
  11. 2026You Don't Need All That Attention: Surgical Memorization Mitigation in Text-to-Image Diffusion Modelscited by 0
  12. 2026Memorization In Stable Diffusion Is Unexpectedly Driven by CLIP Embeddingscited by 0
8 more
  1. 2026Machine Unlearning for Masked Diffusion Language Modelscited by 0
  2. 2026Broken Memories: Detecting and Mitigating Memorization in Diffusion Models with Degraded Generationscited by 0
  3. 2026Localizing Memorized Regions in Diffusion Models via Coordinate-Wise Curvature Differencescited by 0
  4. 2026Initialization is Half the Battle: Generating Diverse Images from a Guidance Potential Posteriorcited by 0
  5. 2026Semi-Supervised Conditional Diffusion via Label Augmentationcited by 0
  6. 2026Fingerprinting Text-to-Image Diffusion Models via Collapsed Generationcited by 0
  7. 2026Background Gradients Shape Memorization in Flow Matchingcited by 0
  8. 2026From Modes to Memories: Characterizing the Scale-Space Dynamics of Diffusion Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Losing dimensions: Geometric memorization in generative diffusion
  2. 2025Does Generation Require Memorization? Creative Diffusion Models using Ambient Diffusion
  3. 2025SolidMark: Evaluating Image Memorization in Generative Models
  4. 2025Provable Separations between Memorization and Generalization in Diffusion Models
  5. 2026The Null Is the Hard Part: Exact Tests for Memorization in Generative Models
  6. 2026Two Calm Ends and the Wild Middle: A Geometric Picture of Memorization in Diffusion Models
  7. 2026Memorization Control in Diffusion Models from Denoising-centric Perspective
  8. 2025Memorization and Regularization in Generative Diffusion Models