Research map / Image generation / Fast diffusion sampling
The Spacetime of Diffusion Models: An Information Geometry Perspective
Rafal Karczewski, Markus Heinonen, Alison Pouplin, Søren Hauberg, Vikas K Garg
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use spacetime geodesic paths instead of MCMC/Doob baselines | transition path sampling · Alanine dipeptide | MaxEnergy: 42.54 -> 37.36 (vs MCMC-fixed-length) | Improved |
| use spacetime geodesic instead of MCMC-fixed-length | transition path sampling · Alanine dipeptide | energy evaluations: 1.29B -> 16M | Improved |
| add SNR penalty (low-variance) or KL region-avoidance penalty to geodesic (ablation) | spacetime geodesic · Alanine dipeptide | path variance / region avoidance | Improved |
| compare geodesics with PF-ODE trajectories | EDM2 · ImageNet-512 | perceptual difference of trajectories | No clear effect |
| compare DiffED with LPIPS and SSIM | DiffED · ImageNet image pairs | correlation: LPIPS -7%, SSIM 53% | Mixed |
Closest papers by meaning
- Hessian Geometry of Latent Space in Generative Models (ICML 2025 · similarity 0.89)
- Adapting to Unknown Low-Dimensional Structures in Score-Based Diffusion Models (NeurIPS 2024 · similarity 0.89)
- Diffusion Model's Generalization Can Be Characterized by Inductive Biases toward a Data-Dependent Ridge Manifold (NeurIPS 2026 · similarity 0.88)
- The Geometry of Noise: Why Diffusion Models Don't Need Noise Conditioning (NeurIPS 2026 · similarity 0.88)
- Conservation Laws for Diffusion Models (NeurIPS 2026 · similarity 0.88)
- Manifold Diffusion Fields (ICLR 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.
