Research map / Image generation / Fast diffusion sampling
Which Layer Causes Distribution Deviation? Entropy-Guided Adaptive Pruning for Diffusion and Flow Models
Changlin Li, Jiawei Zhang, Zeyi Shi, Zhihui Li, Xiaojun Chang
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 CED-guided block ranking (ablation) | EntPruner · Flowers (SiT ODE, 35%) | FID: 12.06 -> 11.75 | Improved |
| add CED-guided block ranking (ablation) | EntPruner · Flowers (SiT ODE, 35%) | IS: 3.81 -> 3.82 | No clear effect |
| add adaptive pruning schedule (ablation) | EntPruner · Flowers (SiT ODE, 35%) | FID: 11.84 -> 11.75 | Improved |
| add adaptive pruning schedule (ablation) | EntPruner · Flowers (SiT ODE, 35%) | IS: 3.80 -> 3.82 | Improved |
| prune blocks with highest CED instead of lowest (ablation) | EntPruner · SiT downstream | FID | Improved |
| replace LD-Pruner with EntPruner at 35% pruning | SiT-XL/2 ODE · ArtBench | FID: 10.78 -> 10.03 | Improved |
| replace LD-Pruner with EntPruner at 30% pruning | SiT-XL/2 · ImageNet 256x256 | FID: 6.81 -> 3.53 | Improved |
| prune 30% with EntPruner vs full fine-tuning | DiT-XL/2 · Flowers | FID: 21.05 -> 11.99 | Improved |
| prune 35% with EntPruner vs full fine-tuning | SiT-XL/2 ODE · Flowers | FID: 11.78 -> 11.75 | Improved |
| prune 30% with EntPruner vs unpruned SiT | SiT-XL/2 · ImageNet 256x256 | FID: 2.15 -> 3.53 | Hurt |
| increase pruning from 35% to 50% with EntPruner | SiT-XL/2 ODE · CUB | FID: 5.48 -> 6.68 | Hurt |
| prune 35% with EntPruner vs fine-tuning | SiT-XL/2 ODE · ArtBench | FID: 8.80 -> 10.03 | Hurt |
Closest papers by meaning
- ALTER: All-in-One Layer Pruning and Temporal Expert Routing for Efficient Diffusion Generation (NeurIPS 2025 · similarity 0.91)
- Importance-Aware OBS Pruning for Diffusion Models (NeurIPS 2026 · similarity 0.89)
- DiP-GO: A Diffusion Pruner via Few-step Gradient Optimization (NeurIPS 2024 · similarity 0.88)
- Faster Diffusion: Rethinking the Role of the Encoder for Diffusion Model Inference (NeurIPS 2024 · similarity 0.87)
- Attention-Driven Training-Free Efficiency Enhancement of Diffusion Models (CVPR 2024 · similarity 0.87)
- Entropic Time Schedulers for Generative Diffusion Models (NeurIPS 2025 · similarity 0.87)
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.
