A theory of learning data statistics in diffusion models, from easy to hard

ArXi:2603.12901v1 Announce Type: cross While diffusion models have emerged as a powerful class of generative models, their learning dynamics remain poorly understood. We address this issue first by empirically showing that standard diffusion models trained on natural images exhibit a distributional simplicity bias, learning simple, pair-wise input statistics before specializing to higher-order correlations.