Research map / Image generation / Fast diffusion sampling
Diffusion Models as Plug-and-Play Priors
Alexandros Graikos, Nikolay Malkin, Nebojsa Jojic, Dimitris Samaras
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 |
|---|---|---|---|
| increase DDPM inference steps from 4 to 256 (ablation) | DDPM TSP solver · TSP N=50 | optimality gap (%): 2.86 -> 1.28 | Improved |
| increase DDPM inference steps from 16 to 256 (ablation) | DDPM TSP solver · TSP N=100 | optimality gap (%): 3.78 -> 2.19 | Improved |
| increase DDPM inference steps from 4 to 16 (ablation) | DDPM TSP solver · TSP N=50 | optimality gap (%): 2.86 -> 2.98 | Mixed |
| train on 20-50 vertex instances instead of 50 only (ablation) | DDPM TSP solver · TSP N=100 | optimality gap (%): 2.19 -> 2.11 | No clear effect |
| use DDPM-inferred tours plus 2-opt vs 2-opt alone | DDPM TSP solver · TSP N=50 | optimality gap (%): 2.95 -> 1.28 | Improved |
| fine-tune U-Net on diffusion-inferred segmentations vs training from scratch on them (ablation) | segmentation network · EnviroAtlas Durham, NC | accuracy (%): 76.0 -> 79.8 | Improved |
| fine-tune U-Net on diffusion-inferred segmentations vs PA supervised | segmentation network · EnviroAtlas Phoenix, AZ | accuracy (%): 6.7 -> 69.6 | Improved |
| fine-tune on diffusion-inferred segmentations vs implicit posterior [34] | segmentation network · EnviroAtlas Phoenix, AZ | accuracy (%): 76.2 -> 69.6 | Hurt |
| fine-tune on diffusion-inferred segmentations vs implicit posterior [34] | segmentation network · EnviroAtlas Durham, NC | class mean IoU (%): 48.4 -> 46.4 | Hurt |
| provide weak labels as input to PA supervised model | U-Net · EnviroAtlas Phoenix, AZ | accuracy (%): 6.7 -> 62.8 | Improved |
Closest papers by meaning
- Denoising Diffusion Bridge Models (ICLR 2024 · similarity 0.91)
- Interpreting and Improving Diffusion Models from an Optimization Perspective (ICML 2024 · similarity 0.90)
- A Mixture-Based Framework for Guiding Diffusion Models (ICML 2025 · similarity 0.90)
- Training Diffusion-based Generative Models with Limited Data (ICML 2025 · similarity 0.90)
- Constrained Diffusion Models via Dual Training (NeurIPS 2024 · similarity 0.90)
- Assessing the quality of denoising diffusion models in Wasserstein distance: noisy score and optimal bounds (NeurIPS 2025 · similarity 0.90)
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.
