Beyond Sample Copying: Structural Memorization in Diffusion Models

arXiv:2603.13419 · cs.LG · Submitted 2026-03-12 · Read on arXiv

cs.LG

Submitted: 2026-03-12

Updated: 2026-10-08

Terminology

• Diffusion Model — 6× in this paper · explained in 9 episodes
• Diffusion Models — 4× in this paper · explained in 6 episodes
• Generalization — 4× in this paper · explained in 13 episodes
• Trace — 1× in this paper · explained in 2 episodes

Sources

• Losing dimensions: Geometric memorization in generative diffusion
• Guiding a diffusion model using sliding windows
• Self-Rectifying Diffusion Sampling with Perturbed-Attention Guidance
• On Inductive Biases That Enable Generalization of Diffusion Transformers
• Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets
• Why Diffusion Models Don't Memorize: The Role of Implicit Dynamical Regularization in Training
• On the Edge of Memorization in Diffusion Models
• Extracting Training Data from Diffusion Models
• On the Interpolation Effect of Score Smoothing in Diffusion Models
• Diffusion Models Beat GANs on Image Synthesis
• Two Calm Ends and the Wild Middle: A Geometric Picture of Memorization in Diffusion Models
• LCMem: A Universal Model for Robust Image Memorization Detection
• Diffusion Models and the Manifold Hypothesis: Log-Domain Smoothing is Geometry Adaptive
• Implicit Regularisation in Diffusion Models: An Algorithm-Dependent Generalisation Analysis
• Bigger Isn't Always Memorizing: Early Stopping Overparameterized Diffusion Models
• How Do Flow Matching Models Memorize and Generalize in Sample Data Subspaces?
• Biased Generalization in Diffusion Models
• Denoising Score Matching with Random Features: Insights on Diffusion Models from Precise Learning Curves
• On Memorization in Diffusion Models
• A solvable generative model with a linear, one-step denoiser

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks