Research map / Image generation / Fast diffusion sampling
First Hitting Diffusion Models for Generating Manifold, Graph and Categorical Data
Mao Ye, Lemeng Wu, qiang liu
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 FHDM instead of DPM | ShapeNet Chair point-cloud generation · ShapeNet Chair | MMD: 12.276 -> 6.644 | Improved |
| Use FHDM instead of DPM | ShapeNet Airplane point-cloud generation · ShapeNet Airplane | MMD: 3.276 -> 3.350 | Hurt |
| Use FHDM instead of Riemannian Score-based | Volcano event generation on the sphere · Volcano | negative log-likelihood: -5.56\pm0.26 -> -1.25\pm0.18 | Hurt |
| Use FHDM instead of Riemannian Score-based | Earthquake event generation on the sphere · Earthquake | negative log-likelihood: -0.21\pm0.03 -> -0.27\pm0.02 | Improved |
| Use FHDM instead of Riemannian Score-based | Flood event generation on the sphere · Flood | negative log-likelihood: 0.52\pm0.02 -> 0.29\pm0.03 | Improved |
| Use FHDM instead of Riemannian Score-based | Fire event generation on the sphere · Fire | negative log-likelihood: -1.24\pm0.07 -> -1.24\pm0.08 | No clear effect |
| Use FHDM instead of EDP-GNN | Community-small graph generation · Community-small | Avg. MMD: 0.074 -> 0.041 | Improved |
| Use FHDM instead of EDP-GNN | Ego-small graph generation · Ego-small | Avg. MMD: 0.050 -> 0.021 | Improved |
| Use FHDM instead of EDP-GNN | Graph generation · Community-small and Ego-small | Avg. MMD: 0.062 -> 0.031 | Improved |
| Use FHDM instead of Argmax / Softplus thresholding | Cityscapes segmentation-map generation · Cityscapes segmentation maps | ELBO: 0.303 -> 0.066 | Improved |
| Use FHDM instead of Argmax / Softplus thresholding | Cityscapes segmentation-map generation · Cityscapes segmentation maps | IWBO: 0.290 -> 0.065 | Improved |
| Reduce EDP-GNN diffusion steps from 4K to 120 (ablation) | EDP-GNN graph generation · Community-small and Ego-small | Avg. MMD: 0.062 -> 0.306 | Hurt |
Closest papers by meaning
- Accelerating Diffusion Models with Parallel Sampling: Inference at Sub-Linear Time Complexity (NeurIPS 2024 · similarity 0.89)
- Dimension-free Score Matching and Time Bootstrapping for Diffusion Models (NeurIPS 2025 · similarity 0.89)
- One-Step Diffusion Distillation through Score Implicit Matching (NeurIPS 2024 · similarity 0.88)
- Diffusion Model's Generalization Can Be Characterized by Inductive Biases toward a Data-Dependent Ridge Manifold (NeurIPS 2026 · similarity 0.88)
- Accelerating Convergence of Score-Based Diffusion Models, Provably (ICML 2024 · similarity 0.88)
- Adapting to Unknown Low-Dimensional Structures in Score-Based Diffusion Models (NeurIPS 2024 · similarity 0.88)
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.
