Research map / Image generation / Fast diffusion sampling
Test-Time Guidance for Flow-Based Generative Models via Parallel Tempering on Source Distributions
Shih-Hsin Wang, Joel Keller, Taos Transue, Drake Benjamin Brown, Thomas Strohmer, Bao Wang
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 SPT instead of FK Steering | Stable Diffusion 1.5 · ImageReward benchmark | ImageReward: 0.898 -> 1.411 | Improved |
| use SPT instead of FK Steering | Stable Diffusion 1.4 · ImageReward benchmark | ImageReward: 0.927 -> 1.409 | Improved |
| use SPT instead of FK Steering | Stable Diffusion XL · ImageReward benchmark | ImageReward: 1.298 -> 1.479 | Improved |
| use SPT instead of FK Steering | Stable Diffusion 1.4 · ImageReward prompts | CLIP diversity: 0.193 -> 0.346 | Improved |
| use SPT instead of FK Steering | Stable Diffusion XL · ImageReward prompts | CLIP diversity: 0.123 -> 0.351 | Improved |
| use SPT instead of unsteered base model | Stable Diffusion 1.4 · ImageReward prompts | CLIP diversity: 0.316 -> 0.346 | Improved |
| use SPT instead of FK Steering at equal 5,250 NFE | Stable Diffusion · ImageReward benchmark | ImageReward (median): p-value 1.3 x 10^-11 | Improved |
| use SPT instead of FK Steering | flow matching model · Lorenz '63 steering | success rate: 0.94 -> 0.99 | Improved |
| use SPT instead of FK Steering | flow matching model · Lorenz '63 steering | total variation distance: 0.13 -> 0.14 | No clear effect |
| use SPT instead of FK Steering | flow matching model · FitzHugh-Nagumo steering | success rate: 0.35 -> 0.81 | Improved |
| use SPT instead of FK Steering | flow matching model · FitzHugh-Nagumo steering | total variation distance: 0.80 -> 0.64 | Improved |
| use SPT instead of FK Steering as the target mode narrows (smaller omega^2) | FK Steering · 1D tilted Gaussian mixture | Wasserstein distance | Improved |
Closest papers by meaning
- Greed is Good: A Unifying Perspective on Guided Generation (NeurIPS 2025 · similarity 0.88)
- REG: Rectified Gradient Guidance for Conditional Diffusion Models (ICML 2025 · similarity 0.87)
- Improved Training Technique for Shortcut Models (NeurIPS 2025 · similarity 0.86)
- Test-time Scaling of Diffusions with Flow Maps (NeurIPS 2026 · similarity 0.86)
- Tortoise and Hare Guidance: Accelerating Diffusion Model Inference with Multirate Integration (NeurIPS 2025 · similarity 0.86)
- Tempered Guided Diffusion (NeurIPS 2026 · similarity 0.86)
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.
