Research map / Image generation / Autoregressive and latent generation
Generative Modelling with Inverse Heat Dissipation
Severi Rissanen, Markus Heinonen, Arno Solin
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 inverse heat dissipation instead of standard diffusion | IHDM · CIFAR-10 | FID: 18.96 vs 3.17 | Hurt |
| Set forward-process noise σ to zero (ablation) | IHDM · MNIST | sample quality | Hurt |
| Set sampling noise δ to zero (ablation) | IHDM · Image generation | image detail | Hurt |
| Increase sampling noise δ above the training noise σ (ablation) | IHDM · Image generation | image detail: 1.25 to 1.3 | Improved |
| Increase sampling noise δ too far (ablation) | IHDM · Image generation | sample quality | Hurt |
| Train on only the first 20 MNIST digits (ablation) | IHDM · MNIST | generalization: 20 | Improved |
Closest papers by meaning
- Warm Diffusion: Recipe for Blur-Noise Mixture Diffusion Models (ICLR 2025 · similarity 0.90)
- Smooth Diffusion: Crafting Smooth Latent Spaces in Diffusion Models (CVPR 2024 · similarity 0.89)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.88)
- Rethinking Peculiar Images by Diffusion Models: Revealing Local Minima’s Role (AAAI 2024 · similarity 0.88)
- Denoising Diffusion Bridge Models (ICLR 2024 · similarity 0.88)
- Color Alignment in Diffusion (CVPR 2025 · 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.
