Research map / Image generation / Fast diffusion sampling
ERTACache: Error Rectification and Timesteps Adjustment for Efficient Diffusion
Xurui Peng, Chenqian Yan, Hong Liu, Rui Ma, Fangmin Chen, XING WANG, Zhihua Wu, Songwei Liu, Mingbao Lin
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 offline-calibrated cache policy instead of uniform cache (ablation) | ERTACache · Wan2.1-1.3B | PSNR: 13.7640 -> 22.0920 | Improved |
| use offline-calibrated cache policy instead of uniform cache (ablation) | ERTACache · Flux-dev 1.0 | PSNR: 15.9428 -> 19.4069 | Improved |
| add trajectory-aware timestep adjustment (ablation) | ERTACache offline policy · Wan2.1-1.3B | PSNR: 22.0920 -> 22.9413 | Improved |
| add trajectory-aware timestep adjustment (ablation) | ERTACache offline policy · Flux-dev 1.0 | PSNR: 19.4069 -> 20.3586 | Improved |
| add explicit error rectification (ablation) | ERTACache with time adjustment · Wan2.1-1.3B | PSNR: 22.9413 -> 23.7680 | Improved |
| add explicit error rectification (ablation) | ERTACache with time adjustment · Wan2.1-1.3B | VBench (%): 80.89% -> 80.73% | Hurt |
| add explicit error rectification (ablation) | ERTACache with time adjustment · Flux-dev 1.0 | PSNR: 20.3586 -> 20.5146 | Improved |
| ERTACache instead of TeaCache at similar speedup | CogVideoX fast · CogVideoX | PSNR: 20.97 -> 26.44 | Improved |
| ERTACache instead of TeaCache | Wan2.1-1.3B · Wan2.1-1.3B | VBench (%): 76.04% -> 80.73% | Improved |
| ERTACache-slow instead of full-step Open-Sora 1.2 | Open-Sora 1.2 · Open-Sora 1.2 | VBench (%): 79.22% -> 79.36% | No clear effect |
| ERTACache instead of FasterCache at same speedup | CogVideoX slow · CogVideoX | VBench (%): 79.83% -> 79.30% | Hurt |
Closest papers by meaning
- Timestep Embedding Tells: It's Time to Cache for Video Diffusion Model (CVPR 2025 · similarity 0.90)
- FasterCache: Training-Free Video Diffusion Model Acceleration with High Quality (ICLR 2025 · similarity 0.90)
- MagCache: Fast Video Generation with Magnitude-Aware Cache (NeurIPS 2025 · similarity 0.89)
- Accelerating Rectified Flow Models via Trajectory-Aware Caching (NeurIPS 2026 · similarity 0.89)
- DeepCache: Accelerating Diffusion Models for Free (CVPR 2024 · similarity 0.89)
- CacheQuant: Comprehensively Accelerated Diffusion Models (CVPR 2025 · similarity 0.89)
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.
