The Null Is the Hard Part: Exact Tests for Memorization in Generative Models

arXiv:2610.00251 · cs.LG · Submitted 2026-09-23 · Read on arXiv

cs.LG

Submitted: 2026-09-23

Updated: 2026-09-23

Terminology

• Generative Model — 1× in this paper · explained in 2 episodes
• Generative Models — 1× in this paper · explained in 3 episodes

Sources

• On Memorization in Diffusion Models
• MemBench: Memorized Image Trigger Prompt Dataset for Diffusion Models
• Instance-Level Data-Use Auditing of Visual ML Models
• An analytic theory of creativity in convolutional diffusion models
• The Intersection Euler Characteristic Profile: Euler Calculus and Stability for Topological Interaction of Ball Unions
• Certified Topological Interaction in Neural Representations: Exact Tests and the Statistic They Require
• Scalable Extraction of Training Data from (Production) Language Models
• A Geometric Framework for Understanding Memorization in Generative Models
• A Reproducible Extraction of Training Images from Diffusion Models

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