Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.15704 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910021813534720 |
|---|---|
| author | Liang, Yuxuan Li, Xu Chen, Xiaolei Zheng, Yi Chen, Haotian Li, Bin Xue, Xiangyang |
| author_facet | Liang, Yuxuan Li, Xu Chen, Xiaolei Zheng, Yi Chen, Haotian Li, Bin Xue, Xiangyang |
| contents | Large Vision-Language Models (LVLMs) have recently demonstrated strong multimodal understanding, yet their fine-grained visual perception is often constrained by low input resolutions. A common remedy is to partition high-resolution images into multiple sub-images for separate encoding, but this approach drastically inflates the number of visual tokens and introduces prohibitive inference overhead. To overcome this challenge, we propose Pyramid Token Pruning (PTP), a training-free strategy that hierarchically integrates bottom-up visual saliency at both region and token levels with top-down instruction-guided relevance. Inspired by human visual cognition, PTP selectively preserves more tokens from salient regions while further emphasizing those most relevant to task instructions. Extensive experiments on 13 diverse benchmarks show that PTP substantially reduces computational cost, memory usage, and inference latency, with negligible performance degradation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_15704 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Pyramid Token Pruning for High-Resolution Large Vision-Language Models via Region, Token, and Instruction-Guided Importance Liang, Yuxuan Li, Xu Chen, Xiaolei Zheng, Yi Chen, Haotian Li, Bin Xue, Xiangyang Computer Vision and Pattern Recognition Large Vision-Language Models (LVLMs) have recently demonstrated strong multimodal understanding, yet their fine-grained visual perception is often constrained by low input resolutions. A common remedy is to partition high-resolution images into multiple sub-images for separate encoding, but this approach drastically inflates the number of visual tokens and introduces prohibitive inference overhead. To overcome this challenge, we propose Pyramid Token Pruning (PTP), a training-free strategy that hierarchically integrates bottom-up visual saliency at both region and token levels with top-down instruction-guided relevance. Inspired by human visual cognition, PTP selectively preserves more tokens from salient regions while further emphasizing those most relevant to task instructions. Extensive experiments on 13 diverse benchmarks show that PTP substantially reduces computational cost, memory usage, and inference latency, with negligible performance degradation. |
| title | Pyramid Token Pruning for High-Resolution Large Vision-Language Models via Region, Token, and Instruction-Guided Importance |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2509.15704 |