The critical slowing down in diffusion models

• URL: http://arxiv.org/abs/2605.12597v2
• Date: Wed, 20 May 2026 15:46:04 GMT
• Title: The critical slowing down in diffusion models
• Abstract summary: We show that training a score model with a one-layer network architecture matching the exact solution exhibits a form of critical slowing down in parameter learning.<n>This slowing down also impacts the generation process, indicating that the well-known difficulties of sampling near criticality persist even for learned generative models.<n>We find that using a two-layer architecture drastically reduces the critical slowing down, with the training time scaling logarithmically rather than quadratically with system size.
• Score: 8.207196072624464
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Computational sampling has been central to the sciences since the mid-20th century. While machine-learning-based approaches have recently enabled major advances, their behavior remains poorly understood, with limited theoretical control over when and why they succeed. Here we provide such insight for diffusion models-a class of generative schemes highly effective in practice-by analyzing their application to the $O(n)$ model of statistical field theory in the Gaussian limit $n \to \infty$. In this analytically tractable setting, we show that training a score model with a one-layer network architecture matching the exact solution exhibits a form of critical slowing down in parameter learning. This slowing down also impacts the generation process, indicating that the well-known difficulties of sampling near criticality persist even for learned generative models. To overcome this bottleneck, we demonstrate the power of combining architectural depth with physical locality. We find that using a two-layer architecture drastically reduces the critical slowing down, with the training time scaling logarithmically rather than quadratically with system size. By introducing a local score approximation we show that this acceleration in training time can be achieved without increasing the number of neural network parameters. Taken together, these results demonstrate that diffusion models can overcome the critical slowing down through appropriate architectural design, and establish a controlled framework for understanding and improving learned sampling methods in statistical physics and beyond.

Related papers

• On Neural Scaling Laws for Weather Emulation through Continual Training [76.70310203875866]
  We study neural scaling in Scientific Machine Learning, focusing on models for weather forecasting.<n>To analyze scaling behavior in as simple a setting as possible, we adopt a minimal, scalable, general-purpose Swin Transformer architecture.<n>We show that models trained in this minimalist way follow predictable scaling trends and even outperform standard cosine learning rate schedules.
  arXiv  Detail & Related papers  (2026-03-26T17:37:25Z)
• Efficient Conditional Generation on Scale-based Visual Autoregressive Models [26.81493253536486]
  Efficient Control Model (ECM) is a plug-and-play framework featuring a lightweight control module that introduces control signals via a distributed architecture.<n> ECM refines conditional features using real-time generated tokens, and a shared feed-forward network (FFN) designed to maximize the utilization of its limited capacity.<n>Our method achieves high-fidelity and diverse control over image generation, surpassing existing baselines while significantly improving both training and inference efficiency.
  arXiv  Detail & Related papers  (2025-10-07T06:27:03Z)
• Certified Neural Approximations of Nonlinear Dynamics [51.01318247729693]
  In safety-critical contexts, the use of neural approximations requires formal bounds on their closeness to the underlying system.<n>We propose a novel, adaptive, and parallelizable verification method based on certified first-order models.
  arXiv  Detail & Related papers  (2025-05-21T13:22:20Z)
• Architecture-Aware Learning Curve Extrapolation via Graph Ordinary Differential Equation [33.63030304318472]
  We propose a novel architecture-aware neural differential equation model to forecast learning curves continuously.<n>Our model outperforms current state-of-the-art learning curve methods and extrapolation approaches for both pure time-series modeling and CNN-based learning curves.
  arXiv  Detail & Related papers  (2024-12-20T04:28:02Z)
• Learning Spatiotemporal Dynamical Systems from Point Process Observations [7.381752536547389]
  Current neural network-based model approaches fall short when faced with data that is collected randomly over time and space.<n>In response, we developed a new method that can effectively learn from such process observations.<n>Our model integrates techniques from neural differential equations, neural point processes, implicit neural representations and amortized variational inference.
  arXiv  Detail & Related papers  (2024-06-01T09:03:32Z)
• Adv-KD: Adversarial Knowledge Distillation for Faster Diffusion Sampling [2.91204440475204]
  Diffusion Probabilistic Models (DPMs) have emerged as a powerful class of deep generative models. They rely on sequential denoising steps during sample generation. We propose a novel method that integrates denoising phases directly into the model's architecture.
  arXiv  Detail & Related papers  (2024-05-31T08:19:44Z)
• A Dynamical Model of Neural Scaling Laws [79.59705237659547]
  We analyze a random feature model trained with gradient descent as a solvable model of network training and generalization. Our theory shows how the gap between training and test loss can gradually build up over time due to repeated reuse of data.
  arXiv  Detail & Related papers  (2024-02-02T01:41:38Z)
• Towards a Better Theoretical Understanding of Independent Subnetwork Training [56.24689348875711]
  We take a closer theoretical look at Independent Subnetwork Training (IST) IST is a recently proposed and highly effective technique for solving the aforementioned problems. We identify fundamental differences between IST and alternative approaches, such as distributed methods with compressed communication.
  arXiv  Detail & Related papers  (2023-06-28T18:14:22Z)
• Improving robustness of jet tagging algorithms with adversarial training [56.79800815519762]
  We investigate the vulnerability of flavor tagging algorithms via application of adversarial attacks. We present an adversarial training strategy that mitigates the impact of such simulated attacks.
  arXiv  Detail & Related papers  (2022-03-25T19:57:19Z)
• Continual Learning Beyond a Single Model [28.130513524601145]
  We show that employing ensemble models can be a simple yet effective method to improve continual performance. We propose a computationally cheap algorithm with similar runtime to a single model yet enjoying the performance benefits of ensembles.
  arXiv  Detail & Related papers  (2022-02-20T14:30:39Z)
• Closed-form Continuous-Depth Models [99.40335716948101]
  Continuous-depth neural models rely on advanced numerical differential equation solvers. We present a new family of models, termed Closed-form Continuous-depth (CfC) networks, that are simple to describe and at least one order of magnitude faster.
  arXiv  Detail & Related papers  (2021-06-25T22:08:51Z)
• Balancing Accuracy and Latency in Multipath Neural Networks [0.09668407688201358]
  We use a one-shot neural architecture search model to implicitly evaluate the performance of an intractable number of neural networks. We show that our method can accurately model the relative performance between models with different latencies and predict the performance of unseen models with good precision across different datasets.
  arXiv  Detail & Related papers  (2021-04-25T00:05:48Z)
• An Ode to an ODE [78.97367880223254]
  We present a new paradigm for Neural ODE algorithms, called ODEtoODE, where time-dependent parameters of the main flow evolve according to a matrix flow on the group O(d) This nested system of two flows provides stability and effectiveness of training and provably solves the gradient vanishing-explosion problem.
  arXiv  Detail & Related papers  (2020-06-19T22:05:19Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.