Invited talk MathSTIC · USPN, France
Optimisation et Apprentissage appliqués aux contenus numériques
Invited talk at USPN on Understanding diffusion models requires rethinking (again) generalization.
Slides
The paper behind this talk
Understanding diffusion models requires rethinking (again) generalization
NeurIPS 2026 position track
- Problem. In diffusion models memorization and generalization are incompatible, so the supervised-learning theory of generalization does not transfer.
- Approach. Survey the three families of explanations and run controlled CIFAR-10 sweeps over dataset size, model size, batch size and learning rate.
- Position. Why models do not memorize is largely settled by early stopping and the linear scaling of memorization time; what is learned before memorization is the open question.
