Research map / Image generation / Fast diffusion sampling
Metropolis-Adjusted Diffusion Models
Kevin H. Lam, Tyler Farghly, Christopher Williams, Jun Yang, Yee Whye Teh, Arnaud Doucet
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes Metropolis-adjusted Langevin correctors for diffusion models (MADM) using only the score, via an exact two-coin Bernoulli factory Barker correction and a Simpson's-rule MH approximation, showing gains in FID on CIFAR-10, FFHQ, AFHQv2 and ImageNet-64.
Problem
placeholder
Approach
placeholder
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add one Simpson-rule MADM corrector step per predictor step | PF-ODE with Heun solver · CIFAR-10 | FID: 1.96 -> 1.95 | Improved |
| add one Simpson-rule MADM corrector step per predictor step | PF-ODE with Heun solver · FFHQ | FID: 2.47 -> 2.43 | Improved |
| add one Simpson-rule MADM corrector step per predictor step | PF-ODE with Heun solver · AFHQv2 | FID: 2.04 -> 2.03 | Improved |
| add one Simpson-rule MADM corrector step per predictor step | PF-ODE with Heun solver · ImageNet-64 | FID: 2.32 -> 2.15 | Improved |
| add one Simpson-rule MADM corrector step per predictor step | PF-ODE with Euler solver · CIFAR-10 | FID: 7.62 -> 7.52 | Improved |
| add one Simpson-rule MADM corrector step per predictor step | PF-ODE with Euler solver · FFHQ | FID: 4.59 -> 4.52 | Improved |
| add one Simpson-rule MADM corrector step per predictor step | PF-ODE with Euler solver · AFHQv2 | FID: 2.93 -> 2.91 | Improved |
| add one Simpson-rule MADM corrector step per predictor step | PF-ODE with Euler solver · ImageNet-64 | FID: 2.41 -> 2.25 | Improved |
| Euler predictor with MADM instead of higher-order Heun predictor without corrector | PF-ODE · ImageNet-64 | FID: 2.32 -> 2.25 | Improved |
Closest papers by meaning
- Diffusion Path Samplers via Sequential Monte Carlo (NeurIPS 2026 · similarity 0.89)
- O(d/T) Convergence Theory for Diffusion Probabilistic Models under Minimal Assumptions (ICLR 2025 · similarity 0.88)
- Theory on Score-Mismatched Diffusion Models and Zero-Shot Conditional Samplers (ICLR 2025 · similarity 0.88)
- Sequential Controlled Langevin Diffusions (ICLR 2025 · similarity 0.87)
- Reverse Diffusion Sequential Monte Carlo Samplers (NeurIPS 2025 · similarity 0.87)
- The Poisson Midpoint Method for Langevin Dynamics: Provably Efficient Discretization for Diffusion Models (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
