Research map / Video / Video generation
QuantCache: Adaptive Importance-Guided Quantization with Hierarchical Latent and Layer Caching for Video Generation
Junyi Wu, Zhiteng Li, Zheng Hui, Yulun Zhang, Linghe Kong, Xiaokang Yang
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 hierarchical latent caching (HLC) (ablation) | QuantCache on Open-Sora · Open-Sora prompts | speedup: 1.00x -> 4.12x | Improved |
| add hierarchical latent caching (HLC) (ablation) | QuantCache on Open-Sora · Open-Sora prompts | aesthetic quality: 63.09 -> 62.09 | Hurt |
| add adaptive importance-guided quantization (AIGQ) (ablation) | QuantCache on Open-Sora · Open-Sora prompts | speedup: 4.12x -> 6.33x | Improved |
| add adaptive importance-guided quantization (AIGQ) (ablation) | QuantCache on Open-Sora · Open-Sora prompts | imaging quality: 58.28 -> 55.68 | Hurt |
| add structural redundancy-aware pruning (SRAP) (ablation) | QuantCache on Open-Sora · Open-Sora prompts | speedup: 6.33x -> 6.72x | Improved |
| add structural redundancy-aware pruning (SRAP) (ablation) | QuantCache on Open-Sora · Open-Sora prompts | background consistency: 97.62 -> 96.19 | Hurt |
| QuantCache versus AdaCache-fast | QuantCache · Open-Sora speedup | speedup: 2.24x -> 6.72x | Improved |
| QuantCache 4/6 versus ViDiT-Q 4/8 | QuantCache · VBench | aesthetic quality: 58.23 -> 58.63 | Improved |
| QuantCache 4/6 versus ViDiT-Q 4/8 | QuantCache · VBench | imaging quality: 57.21 -> 55.94 | Hurt |
| lower weight/activation bit-width for Q-DiT (8/8 to 4/8) | Q-DiT · VBench | aesthetic quality: 58.68 -> 31.14 | Hurt |
| QuantCache versus ViDiT-Q at 8/8 | QuantCache · VBench | aesthetic quality: 59.89 -> 58.57 | Hurt |
| QuantCache 4/6 versus ViDiT-Q 4/8 | QuantCache · DOVER | VQA technical: 49.11 -> 49.14 | No clear effect |
Closest papers by meaning
- Q-VDiT: Towards Accurate Quantization and Distillation of Video-Generation Diffusion Transformers (ICML 2025 · similarity 0.92)
- ViDiT-Q: Efficient and Accurate Quantization of Diffusion Transformers for Image and Video Generation (ICLR 2025 · similarity 0.91)
- PTQ4DiT: Post-training Quantization for Diffusion Transformers (NeurIPS 2024 · similarity 0.89)
- $\text{S}^2$Q-VDiT: Accurate Quantized Video Diffusion Transformer with Salient Data and Sparse Token Distillation (NeurIPS 2025 · similarity 0.88)
- Q-DiT: Accurate Post-Training Quantization for Diffusion Transformers (CVPR 2025 · similarity 0.88)
- Training-Free Efficient Video Generation via Dynamic Token Carving (NeurIPS 2025 · 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.
