Research map / Image generation / Fast diffusion sampling
Dual-End Consistency Model
linwei dong, Ruoyu Guo, Ge Bai, Zehuan Yuan, Yawei Luo, Changqing Zou
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 flow matching loss (ablation) | CD only · ImageNet 256 1 NFE | FID: 3.08 -> 1.78 | Improved |
| add flow matching loss (ablation) | CD only · ImageNet 256 50 NFE | FID: 6.52 -> 1.78 | Improved |
| add N2N loss (ablation) | FM + CD · ImageNet 256 50 NFE | FID: 1.78 -> 1.26 | Improved |
| add N2N loss (ablation) | FM + CD · ImageNet 256 1 NFE | FID: 1.78 -> 1.70 | Improved |
| add CD loss (ablation) | FM + N2N · ImageNet 256 1 NFE | FID: 441.66 -> 1.70 | Improved |
| add r-value decay (delta 0.1) (ablation) | DE-CM · ImageNet 256 1 NFE | FID: 1.74 -> 1.70 | Improved |
| use lambda 0.5 instead of 0 (ablation) | DE-CM · ImageNet 256 2 NFE | FID: 1.44 -> 1.33 | Improved |
| use gamma 1 instead of 0.5 (ablation) | DE-CM · ImageNet 256 1 NFE | FID: 1.82 -> 1.70 | Improved |
| use uniform-with-N-seg timestep sampling instead of log-norm (ablation) | DE-CM · ImageNet 256 1 NFE | FID: 1.83 -> 1.70 | Improved |
| update N2N loss every 3 iterations instead of every iteration (ablation) | DE-CM · ImageNet 256 4 NFE | FID: 1.77 -> 1.41 | Improved |
| add velocity normalization (ablation) | DE-CM · ImageNet 256 50 NFE | FID: 1.35 -> 1.26 | Improved |
| add velocity normalization (ablation) | DE-CM · ImageNet 256 2 NFE | FID: 1.30 -> 1.33 | Hurt |
Closest papers by meaning
- Truncated Consistency Models (ICLR 2025 · similarity 0.91)
- Simplifying, Stabilizing and Scaling Continuous-time Consistency Models (ICLR 2025 · similarity 0.91)
- SCoT: Unifying Consistency Models and Rectified Flows via Straight-Consistent Trajectories (NeurIPS 2025 · similarity 0.90)
- Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion (ICLR 2024 · similarity 0.89)
- Improved Techniques for Training Consistency Models (ICLR 2024 · similarity 0.89)
- Improved Training Technique for Latent Consistency Models (ICLR 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.
