Understanding diffusion models requires rethinking (again) generalization
Understanding diffusion models requires new theory since memorization and generalization are incompatible, so research should study what models learn before memorizing.
Published 2026Paris Poster Session 5 · Fri, Dec 11, 11:30 AM–1:30 PM local time · Paris Poster HallarXiv ↗OpenReview ↗
Only vote on papers you've read. Sign in with GitHub to vote.
A sharp pivot to pre-memorization dynamics offers a vital new framework for diffusion generalization, though a single CIFAR-10 study leaves open whether that phase scales beyond small benchmarks.
Abstract
This position paper argues that understanding generalization in diffusion models requires fundamentally new theoretical frameworks that go beyond both classical statistical learning theory and the benign overfitting paradigm developed for supervised learning. In diffusion models, unlike in supervised learning, memorization of training data and generalization to novel samples are incompatible: a model that has fully memorized its training set generates copies rather than novel data. Several theoretical explanations for why practical diffusion models nevertheless generalize have been proposed, based on capacity limitations, implicit regularization from optimization, or architectural inductive biases, but their interactions remain unclear. We argue that the field should pivot from explaining why the diffusion models do not memorize to investigating what the model actually learns during pre-memorization phase. To highlight our stance, we conduct empirical study of diffusion models trained on CIFAR-10, and we distill the findings into concrete open questions that we believe are key to improve understanding of generalization in diffusion models.