Research map / Image generation / Fast diffusion sampling
Temporal Score Rescaling for Temperature Sampling in Diffusion and Flow Models
Yanbo Xu, Yu Wu, Sungjae Park, Zhizhuo Zhou, Shubham Tulsiani
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 |
|---|---|---|---|
| add TSR to Euler-ODE sampling | SD3 · LAION Aesthetics 5k | FID: 24.77 -> 22.81 | Improved |
| add TSR to Euler-ODE sampling | SD3 · LAION Aesthetics 5k | CLIP score: 32.82 -> 33.05 | Improved |
| add TSR to DDPM sampling | SD2 · LAION Aesthetics 5k | FID: 22.81 -> 19.75 | Improved |
| add TSR to default scheduler | Flux.1 dev · LAION Aesthetics 5k | FID: 53.99 -> 51.79 | Improved |
| add TSR to DDIM | Marigold · ETH3D depth | AbsRel: 7.1 -> 6.68 | Improved |
| add TSR to DDIM | Marigold · NYUv2 depth | AbsRel: 6.0 -> 5.84 | Improved |
| add TSR instead of CNS to DDIM | Marigold · ETH3D depth | AbsRel: 6.82 -> 6.68 | Improved |
| add TSR to score sampling | SO(3) diffusion · SYMSOL pose | mean error (deg): 0.444 -> 0.356 | Improved |
| use CNS (1600) instead of TSR | SO(3) diffusion · SYMSOL pose | mean error (deg): 0.356 -> 0.350 | No clear effect |
| add TSR to Pi-0 policy | Pi-0 · LIBERO-10 | success rate: 81.7 -> 82.8 | Improved |
| add TSR with shared k>1 across tasks | Pi-0 · LIBERO-10 Task 8 | success rate: 23.3 -> 21.3 | Hurt |
| add TSR to DDPM instead of constant noise scaling | SD2 · LAION Aesthetics 5k | FID: 19.87 -> 19.57 | Improved |
Closest papers by meaning
- A Simple Early Exiting Framework for Accelerated Sampling in Diffusion Models (ICML 2024 · similarity 0.90)
- CCS: Controllable and Constrained Sampling with Diffusion Models via Initial Noise Perturbation (NeurIPS 2025 · similarity 0.90)
- Align Your Steps: Optimizing Sampling Schedules in Diffusion Models (ICML 2024 · similarity 0.90)
- Test-time Scaling of Diffusions with Flow Maps (NeurIPS 2026 · similarity 0.90)
- Particle Denoising Diffusion Sampler (ICML 2024 · similarity 0.90)
- Denoising Diffusion Bridge Models (ICLR 2024 · 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.
