Model Releases
Importance-Aware OBS Pruning for Diffusion Models
arXiv:2607.20048v1 Announce Type: new Abstract: We propose importance-aware pruning for diffusion models, a training-free framework that prioritizes preserving parameters critical to semantically sali
arXiv:2607.20048v1 Announce Type: new Abstract: We propose importance-aware pruning for diffusion models, a training-free framework that prioritizes preserving parameters critical to semantically salient image regions. To do so, we incorporate spatial importance maps -- derived from conditioning signals or model attention -- into the pruning objective. This produces parameter rankings aligned with perceptual relevance rather than uniform reconstruction error. On MS-COCO dataset, our proposed approach consistently retains subject fidelity and structural correctness at high compression ratios where conventional pruning causes visible degradation. These results demonstrate that content-aware objectives are key to perceptually faithful compression of generative models.
Related
- Training-Free Object-Background Compositional T2I via Dynamic Spatial Guidance and Multi-Path Pruning
- SSR-Merge: Subspace Signal Routing for Training-Free LoRA Merging in Diffusion Models
- SpecEdit: Training-Free Acceleration for Diffusion based Image Editing via Semantic Locking
Source: arXiv cs.CV | 2026-07-23