Concurrence of Symmetry Breaking and Nonlocality Phase Transitions in Diffusion Models

TL;DR

This paper investigates the simultaneous occurrence of symmetry breaking and nonlocality phase transitions in diffusion models, providing a unified diagnostic for understanding their critical dynamics during generation.

Contribution

It is the first to empirically unify the notions of symmetry breaking and nonlocality phase transitions in diffusion models, offering insights for improving efficiency and design.

Findings

Near-simultaneous occurrence of critical times for both phase transitions

Provides a diagnostic for when diffusion models rely on conditioning and global denoising

Guides architecture and sampling scheme design to reduce unnecessary computation

Abstract

Diffusion models undergo a phase transition in a critical time window during generation dynamics, with two complementary diagnoses of criticality. The symmetry breaking picture views the critical window as when trajectories bifurcate into different semantic minima of the energy landscape, whereas the nonlocality picture views the critical window as when local denoising fails. We study whether two notions of such phase transitions are concurrent in modern diffusion transformers. By evaluating the dynamics and outcomes of the generation trajectory, we observe a near-simultaneous occurrence of the non-locality and symmetry breaking critical times. Our work is the first to unify the two notions of phase transitions in practice: it provides a concrete diagnostic for when and why diffusion models rely on conditioning and global denoising, enabling principled evaluation of model efficiency and guiding the design of architectures and sampling schemes that avoid unnecessary computation.