Research map / Image generation / Fast diffusion sampling
ResilPhase: Plug-and-Play Phase Mapping and Noise-Resilient Macro-Trajectory Extrapolation for Diffusion Acceleration
Qicheng Zhao, Yu Li, Qi Sun, Zheyu Yan
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 |
|---|---|---|---|
| predict Global Drift instead of layer-wise output (ablation) | Lagrange extrapolator (N=6,O=1) · DrawBench, FLUX.1-dev | ImageReward: 1.0098 -> 1.0150 | Improved |
| predict Global Drift instead of layer-wise output (speed) (ablation) | Lagrange extrapolator (N=6,O=1) · FLUX.1-dev | speed: 4.06x -> 4.96x | Improved |
| add Balanced Mapping to Global Drift (ablation) | Lagrange extrapolator (N=6,O=1) · DrawBench, FLUX.1-dev | ImageReward: 1.0150 -> 1.0258 | Improved |
| add Balanced Mapping to Global Drift (ablation) | Lagrange extrapolator (N=6,O=1) · FLUX.1-dev | PSNR: 28.734 -> 29.536 | Improved |
| add Chebyshev Mapping to Global Drift (ablation) | Lagrange extrapolator (N=6,O=1) · FLUX.1-dev | PSNR: 28.734 -> 28.734 | No clear effect |
| use Balanced instead of Chebyshev mapping (ablation) | Global Drift Lagrange (N=6,O=1) · DrawBench, FLUX.1-dev | ImageReward: 1.0186 -> 1.0258 | Improved |
| add Balanced Mapping to layer-wise prediction (ablation) | Lagrange extrapolator (N=6,O=1) · DrawBench, FLUX.1-dev | ImageReward: 1.0098 -> 1.0347 | Improved |
| add Balanced Mapping to Global Drift (ablation) | Lagrange extrapolator (N=4,O=1) · DrawBench, FLUX.1-dev | ImageReward: 1.0573 -> 1.0647 | Improved |
| add Chebyshev Mapping to Global Drift (ablation) | Lagrange extrapolator (N=4,O=1) · FLUX.1-dev | PSNR: 29.447 -> 29.448 | No clear effect |
| predict Global Drift instead of layer-wise output (ablation) | Lagrange extrapolator (N=5,O=1) · DrawBench, FLUX.1-dev | ImageReward: 1.0404 -> 1.0557 | Improved |
| add Balanced Mapping to Global Drift (ablation) | Lagrange extrapolator (N=5,O=1) · FLUX.1-dev | PSNR: 29.093 -> 29.556 | Improved |
| add ResilPhase (N=3,O=4) in place of full 50-step DDIM | DiT-XL/2 · ImageNet class-conditional | FID: 2.367 -> 2.347 | Improved |
Closest papers by meaning
- Tortoise and Hare Guidance: Accelerating Diffusion Model Inference with Multirate Integration (NeurIPS 2025 · similarity 0.87)
- GeoSPRINT: Geometric Redundancy-Aware Step Pruning for Inference in Diffusion Trajectories (NeurIPS 2026 · similarity 0.87)
- Enhanced Diffusion Sampling via Extrapolation with Multiple ODE Solutions (ICLR 2025 · similarity 0.87)
- DiRecT: Safe Diffusion-Based Planning via Receding-Horizon Denoising (NeurIPS 2026 · similarity 0.86)
- Faster Diffusion: Rethinking the Role of the Encoder for Diffusion Model Inference (NeurIPS 2024 · similarity 0.86)
- MaRS: A Fast Sampler for Mean Reverting Diffusion based on ODE and SDE Solvers (ICLR 2025 · 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.
