Research map / Video / Video generation
SSBP: Stage-Specialized Block Pruning for Video Diffusion Models
Mengmeng Ge, Takashi Isobe, Dong Zhou, Dong Li, Emad Barsoum
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 |
|---|---|---|---|
| partition denoising into stages and prune per stage instead of one global pruning (ablation) | SSBP · VBench 30% acceleration | total score: 78.07 -> 80.91 | Improved |
| add Inter-Stage Variance Selector (ablation) | SSBP · VBench 30% acceleration | total score: 80.91 -> 82.44 | Improved |
| add Intra-Stage Quality Selector (ablation) | SSBP · VBench 30% acceleration | total score: 82.44 -> 83.52 | Improved |
| use 4 stages instead of 1 (ablation) | SSBP · VBench 30% acceleration | total score: 80.16 -> 83.52 | Improved |
| use 4 stages instead of 2 (ablation) | SSBP · VBench 20% acceleration | total score: 82.70 -> 84.02 | Improved |
| use 3 candidates in IQS instead of 1 (ablation) | SSBP · VBench 30% acceleration | total score: 82.44 -> 83.52 | Improved |
| increase IQS candidates from 3 to 5 or 7 (ablation) | SSBP · VBench 30% acceleration | total score: 83.52 -> 83.63 (N=5), 83.60 (N=7) | No clear effect |
| use SSBP instead of TinyFusion | SSBP · VBench 30% acceleration | total score: 80.46 -> 83.52 | Improved |
| use SSBP instead of L-OBS | SSBP · VBench 20% acceleration | total score: 79.62 -> 84.02 | Improved |
| prune 30% per-step cost versus unpruned DMD2 model | SSBP · VBench | total score: 84.21 -> 83.52 | Hurt |
| compare pruned model quality vs semantic score to DMD2 | SSBP · VBench | quality vs semantic score: quality 85.98 -> 85.02; semantic 77.09 -> 77.54 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CogVideoX-5B — CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer (ICLR 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DiP-GO: A Diffusion Pruner via Few-step Gradient Optimization (NeurIPS 2024 · similarity 0.89)
- ALTER: All-in-One Layer Pruning and Temporal Expert Routing for Efficient Diffusion Generation (NeurIPS 2025 · similarity 0.89)
- Attention-Driven Training-Free Efficiency Enhancement of Diffusion Models (CVPR 2024 · similarity 0.89)
- AsymRnR: Video Diffusion Transformers Acceleration with Asymmetric Reduction and Restoration (ICML 2025 · similarity 0.89)
- Spectral Progressive Diffusion for Efficient Image and Video Generation (NeurIPS 2026 · similarity 0.88)
- DSAQuant: Denoising-Stage-Aligned Quantization-Aware Training for Video Generation (NeurIPS 2026 · 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.
