Identifies 'critical times' in diffusion generation where targeted guidance pulses significantly improve image control.
March 23, 2026
Original Paper
How Out-of-Equilibrium Phase Transitions can Seed Pattern Formation in Trained Diffusion Models
arXiv · 2603.20092
The Takeaway
Interprets reverse diffusion through the lens of non-equilibrium physics and phase transitions. It proves that generation isn't a smooth process but passes through critical regimes, allowing practitioners to optimize sampling and guidance timing for better results.
From the abstract
In this work, we propose a theoretical framework that interprets the generation process in trained diffusion models as an instance of out-of-equilibrium phase transitions. We argue that, rather than evolving smoothly from noise to data, reverse diffusion passes through a critical regime in which small spatial fluctuations are amplified and seed the emergence of large-scale structure. Our central insight is that architectural constraints, such as locality, sparsity, and translation equivariance,