mlsanityData to 7 Oct 2026

Optimal Stepsize for Diffusion Sampling

arXiv 2503.21774 ↗First posted 27 Mar 2025cs.CVcited by 5

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

Diffusion models achieve remarkable generation quality but suffer from computational intensive sampling due to suboptimal step discretization. While existing works focus on optimizing denoising directions, we address the principled design of stepsize schedules. This paper proposes Optimal Stepsize Distillation, a dynamic programming framework that extracts theoretically optimal schedules by distilling knowledge from reference trajectories. By reformulating stepsize optimization as recursive error minimization, our method guarantees global discretization bounds through optimal substructure exploitation. Crucially, the distilled schedules demonstrate strong robustness across architectures, ODE solvers, and noise schedules. Experiments show 10x accelerated text-to-image generation while preserving 99.4% performance on GenEval. Our code is available at https://github.com/bebebe666/OptimalSteps.

Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 36 of its 38 references that are in the map. First posted on arXiv 27 Mar 2025.

Builds on

36 of 38 references.

Closest ancestors · 24

Same direction, newest first

  1. 2024Learning to Discretize Denoising Diffusion ODEs24 May
  2. 2024On the Trajectory Regularity of ODE-based Diffusion Sampling18 May
  3. 2024Align Your Steps: Optimizing Sampling Schedules in Diffusion Models22 Apr
  4. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesis5 Mar
  5. 2024Accelerating Diffusion Sampling with Optimized Time Steps27 Feb
  6. 2023Fast ODE-based Sampling for Diffusion Models in Around 5 Steps30 Nov
  7. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis4 Jul
  8. 2023Efficient Diffusion Training via Min-SNR Weighting Strategy16 Mar
  9. 2023Consistency Models2 Mar
  10. 2023UniPC: A Unified Predictor-Corrector Framework for Fast Sampling of Diffusion Models9 Feb
  11. 2022Scalable Diffusion Models with Transformers19 Dec
  12. 2022DPM-Solver++: Fast Solver for Guided Sampling of Diffusion Probabilistic Models2 Nov
12 more
  1. 2022DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps2 Jun
  2. 2022Elucidating the Design Space of Diffusion-Based Generative Models1 Jun
  3. 2022Pseudo Numerical Methods for Diffusion Models on Manifolds20 Feb
  4. 2022Learning Fast Samplers for Diffusion Models by Differentiating Through Sample Quality11 Feb
  5. 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
  6. 2021Learning to Efficiently Sample from Diffusion Probabilistic Models7 Jun
  7. 2021Gotta Go Fast When Generating Data with Score-Based Models28 May
  8. 2021Improved Denoising Diffusion Probabilistic Models18 Feb
  9. 2020Score-Based Generative Modeling through Stochastic Differential Equations26 Nov
  10. 2020Denoising Diffusion Implicit Models6 Oct
  11. 2020Denoising Diffusion Probabilistic Models19 Jun
  12. 2020Improved Techniques for Training Score-Based Generative Models16 Jun

Other references · 12

  1. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  2. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,441
  3. 2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 5,986
  4. 2015Rethinking the Inception Architecture for Computer Visioncited by 5,451
  5. 2022Flow Matching for Generative Modelingcited by 5,426
  6. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flowcited by 3,550
  7. 2015LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loopcited by 1,350
  8. 2023GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignmentcited by 946
  9. 2022eDiff-I: Text-to-Image Diffusion Models with an Ensemble of Expert Denoiserscited by 921
  10. 2024Autoregressive Image Generation without Vector Quantizationcited by 644
  11. 2024Open-Sora: Democratizing Efficient Video Production for Allcited by 637
  12. 2024Improving the Training of Rectified Flowscited by 106

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 3.

5 papers.

  1. 2025ResearchCodeBench: Benchmarking LLMs on Implementing Novel Machine Learning Research Codecited by 34
  2. 2025Toward Theoretical Insights into Diffusion Trajectory Distillation via Operator Mergingcited by 2
  3. 2026VADF: Vision-Adaptive Diffusion Policy Framework for Efficient Robotic Manipulationcited by 0
  4. 2026Enforcing Constraints in Generative Sampling via Adaptive Correction Schedulingcited by 0
  5. 2026Distill Locally, Schedule Globally: Flow Maps for Few-Step Text-to-Speechcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Fewer Denoising Steps or Cheaper Per-Step Inference: Towards Compute-Optimal Diffusion Model Deployment
  2. 2026Efficient Text-to-Image Generation: An Adaptive Step Schedule Controller for Diffusion Models
  3. 2023AdaDiff: Adaptive Step Selection for Fast Diffusion Models
  4. 2023AutoDiffusion: Training-Free Optimization of Time Steps and Architectures for Automated Diffusion Model Acceleration
  5. 2026Adaptive Second-Order Solvers for Fast Stochastic Diffusion Sampling
  6. 2024A Simple Early Exiting Framework for Accelerated Sampling in Diffusion Models
  7. 2023Parallel Sampling of Diffusion Models
  8. 2025F-scheduler: illuminating the free-lunch design space for fast sampling of diffusion models