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.
| Comments: | 20 pages, 10 figures. comments are welcome |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2605.04830 [cs.LG] |
| (or arXiv:2605.04830v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.04830 arXiv-issued DOI via DataCite (pending registration) |
From: Yifan Zhang [view email]
[v1]
Wed, 6 May 2026 12:24:57 UTC (16,295 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。