Research map / Image generation / Fast diffusion sampling
Plug-and-Play Fidelity Optimization for Diffusion Transformer Acceleration via Cumulative Error Minimization
Tong Shao, Yusen Fu, Guoying Sun, Jingde Kong, Zhuotao Tian, Jingyong Su
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 dynamic caching strategy (DCS, dynamic programming) (ablation) | naive fixed-interval caching · PixArt-alpha | FID: 30.04 -> 28.69 | Improved |
| add dynamic caching strategy (DCS, dynamic programming) (ablation) | naive fixed-interval caching · Hunyuan | VBench (%): 77.64 -> 79.21 | Improved |
| add dynamic caching strategy (DCS, dynamic programming) (ablation) | naive fixed-interval caching · DiT-XL/2 | FID: 3.83 -> 2.73 | Improved |
| add dynamic caching strategy (DCS, dynamic programming) (ablation) | naive fixed-interval caching · DiT-XL/2 | IS: 213.12 -> 234.78 | Improved |
| add cumulative error approximation (CEA) to DCS (ablation) | DCS · PixArt-alpha | FID: 28.69 -> 27.94 | Improved |
| add cumulative error approximation (CEA) to DCS (ablation) | DCS · Hunyuan | VBench (%): 79.21 -> 80.44 | Improved |
| add cumulative error approximation (CEA) to DCS (ablation) | DCS · DiT-XL/2 | FID: 2.73 -> 2.65 | Improved |
| add cumulative error approximation (CEA) to DCS (ablation) | DCS · DiT-XL/2 | IS: 234.78 -> 235.11 | Improved |
| plug in CEM cache strategy | FasterSD · Stable Diffusion 1.5 | FID (10K): 21.62 -> 19.99 | Improved |
| plug in CEM cache strategy | DuCa · PixArt-alpha | FID (50K): 41.56 -> 27.57 | Improved |
| plug in CEM cache strategy | TaylorSeer · FLUX.1-dev | ImageReward: 0.9410 -> 0.9811 | Improved |
| plug in CEM cache strategy | TeaCache · FLUX.1-dev | ImageReward: 0.7228 -> 0.7362 | Improved |
Closest papers by meaning
- Accelerating Diffusion Transformer via Increment-Calibrated Caching with Channel-Aware Singular Value Decomposition (CVPR 2025 · similarity 0.92)
- HarmoniCa: Harmonizing Training and Inference for Better Feature Caching in Diffusion Transformer Acceleration (ICML 2025 · similarity 0.91)
- Learning-to-Cache: Accelerating Diffusion Transformer via Layer Caching (NeurIPS 2024 · similarity 0.89)
- FasterDiT: Towards Faster Diffusion Transformers Training without Architecture Modification (NeurIPS 2024 · similarity 0.89)
- Faster Diffusion: Rethinking the Role of the Encoder for Diffusion Model Inference (NeurIPS 2024 · similarity 0.88)
- FlexiDiT: Your Diffusion Transformer Can Easily Generate High-Quality Samples with Less Compute (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.
