![[2602.15914] Steering Dynamical Regimes of Diffusion Models by Breaking Detailed Balance](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.15914] Steering Dynamical Regimes of Diffusion Models by Breaking Detailed Balance
Summary
This paper explores how breaking detailed balance in generative diffusion processes can enhance reverse processes while maintaining stationary distributions, focusing on the Ornstein-Uhlenbeck process.
Why It Matters
Understanding the dynamics of diffusion models is crucial for advancements in machine learning and statistical mechanics. This research provides insights into optimizing generative models, which can lead to more efficient algorithms and applications in various fields, including AI and data science.
Key Takeaways
- Breaking detailed balance can accelerate reverse processes in diffusion models.
- The study introduces a non-reversible perturbation that optimizes long-time relaxation rates.
- Numerical experiments validate the theoretical findings, enhancing the understanding of phase transitions in generative models.
Condensed Matter > Statistical Mechanics arXiv:2602.15914 (cond-mat) [Submitted on 17 Feb 2026] Title:Steering Dynamical Regimes of Diffusion Models by Breaking Detailed Balance Authors:Haiqi Lu, Ying Tang View a PDF of the paper titled Steering Dynamical Regimes of Diffusion Models by Breaking Detailed Balance, by Haiqi Lu and Ying Tang View PDF HTML (experimental) Abstract:We show that deliberately breaking detailed balance in generative diffusion processes can accelerate the reverse process without changing the stationary distribution. Considering the Ornstein--Uhlenbeck process, we decompose the dynamics into a symmetric component and a non-reversible anti-symmetric component that generates rotational probability currents. We then construct an exponentially optimal non-reversible perturbation that improves the long-time relaxation rate while preserving the stationary target. We analyze how such non-reversible control reshapes the macroscopic dynamical regimes of the phase transitions recently identified in generative diffusion models. We derive a general criterion for the speciation time and show that suitable non-reversible perturbations can accelerate speciation. In contrast, the collapse transition is governed by a trace-controlled phase-space contraction mechanism that is fixed by the symmetric component, and the corresponding collapse time remains unchanged under anti-symmetric perturbations. Numerical experiments on Gaussian mixture models support these findings....
