Low-Dimensional Adaptation of Rectified Flow: A Diffusion and Stochastic Localization Perspective
stat.ML, cs.AI, cs.LG, math.ST, stat.TH
Submitted: 2026-01-21
Updated: 2026-11-18
Code: https://github.com/roysaptaumich/Low-dim-RF
Project page: https://diffusionflow.github.io
Terminology
Sources
- Building Normalizing Flows with Stochastic Interpolants
- Stochastic Interpolants: A Unifying Framework for Flows and Diffusions
- Joint stochastic localization and applications
- Convergence of Diffusion Models Under the Manifold Hypothesis in High-Dimensions
- On the Convergence and Straightness of Rectified Flow
- Nearly $d$-Linear Convergence Bounds for Diffusion Models via Stochastic Localization
- Mirror Flow Matching with Heavy-Tailed Priors for Generative Modeling on Convex Domains
- Denoising diffusion probabilistic models are optimally adaptive to unknown low dimensionality
- FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space
- Towards a mathematical theory for consistency training in diffusion models
- A Sharp Convergence Theory for The Probability Flow ODEs of Diffusion Models
- Low-dimensional adaptation of diffusion models: Convergence in total variation
- Flow Matching for Generative Modeling
- Improving Video Generation with Human Feedback
- Rectified Flow: A Marginal Preserving Approach to Optimal Transport
- Sampling, Diffusions, and Stochastic Localization
- The Intrinsic Dimension of Images and Its Impact on Learning
- Linear Convergence of Diffusion Models Under the Manifold Hypothesis
- Perspectives on Stochastic Localization
- Discriminative Estimation of Total Variation Distance: A Fidelity Auditor for Generative Data
Related papers
- Uniform-in-Time Weak Propagation-of-Chaos in Shallow Neural Networks
- Diffusion Models Observe Only Gradients: A Geometric Perspective on Score Matching Errors
- Quantifying Error Propagation and Model Collapse in Diffusion Models
- Behavior of prediction performance metrics with rare events
- Statistical Inference and Learning for Shapley Additive Explanations (SHAP)
- Control Variate Design for Score-Function Gradient Estimators in Importance-Weighted Variational Inference