Research map / Architectures and efficiency / Attention and transformers
V-Pruner: A Fast and Globally-informed Token Pruning Framework for Vision Transformer
Guangzhen Yao, Jiayun Zheng, Zezhou Wang, Wenxin Zhang, Renda Han, Chuangxin Zhao, Zeyu Zhang, Runhao Liu
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 |
|---|---|---|---|
| use PPO instead of A2C for the pruning policy (ablation) | V-Pruner · ImageNet-1K (ViT-L, 35 GFLOPs) | Top-1 accuracy: 87.45 -> 89.27 | Improved |
| use PPO instead of REINFORCE for the pruning policy (ablation) | V-Pruner · ImageNet-1K (ViT-L, 35 GFLOPs) | Top-1 accuracy: 84.45 -> 89.27 | Improved |
| use learned PPO policy instead of random pruning (ablation) | V-Pruner · ImageNet-1K (DeiT-B, 15 GFLOPs) | Top-1 accuracy: 78.95 -> 85.45 | Improved |
| use PPO instead of A2C (ablation) | V-Pruner · ImageNet-1K (DeiT-B, 15 GFLOPs) | Top-1 accuracy: 84.25 -> 85.45 | Improved |
| use V-Pruner instead of MCTF | DeiT-B at 15 GFLOPs · ImageNet-1K | Top-1 accuracy: 84.16 -> 85.45 | Improved |
| use V-Pruner instead of MCTF | DeiT-B at 15 GFLOPs · ImageNet-1K | training time: 1.3 hours, about 1/6 of MCTF | Improved |
| use V-Pruner instead of training-free ECTP | DeiT-B at 15 GFLOPs · ImageNet-1K | training time: 1.1 hours -> 1.3 hours | Hurt |
| use V-Pruner instead of MCTF on downstream CIFAR-100 | DeiT-B · CIFAR-100 | accuracy: 98.3 -> 98.8 | Improved |
| use V-Pruner instead of MCTF on Pets | DeiT-B · Pets (downstream) | accuracy: 88.5 -> 89.8 | Improved |
| prune tokens with V-Pruner | DeiT-B segmentation · Pascal VOC | mIoU: 79.5 -> 78.6 | Hurt |
Closest papers by meaning
- Exploring Token Pruning in Vision State Space Models (NeurIPS 2024 · similarity 0.90)
- Collaborative Multi-Mode Pruning for Vision-Language Models (CVPR 2026 · similarity 0.90)
- When and What to Prune? Stage-Aware Visual Token Pruning for Efficient VLA (NeurIPS 2026 · similarity 0.89)
- Zero-TPrune: Zero-Shot Token Pruning through Leveraging of the Attention Graph in Pre-Trained Transformers (CVPR 2024 · similarity 0.89)
- Balanced Token Pruning: Accelerating Vision Language Models Beyond Local Optimization (NeurIPS 2025 · similarity 0.89)
- VLM-Pruner: Buffering for Spatial Sparsity in an Efficient VLM Centrifugal Token Pruning Paradigm (CVPR 2026 · 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.
