HieraVid: Hierarchical Token Pruning for Fast Video Large Language Models
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866915909258444800 |
|---|---|
| author | Guo, Yansong Zhu, Chaoyang Ji, Jiayi Lin, Jianghang Cao, Liujuan |
| author_facet | Guo, Yansong Zhu, Chaoyang Ji, Jiayi Lin, Jianghang Cao, Liujuan |
| contents | Video Large Language Models (VideoLLMs) have demonstrated impressive capabilities in video understanding, yet the massive number of input video tokens incurs a significant computational burden for deployment. Existing methods mainly prune video tokens at input level while neglecting the inherent information structure embedded in videos and large language models (LLMs). To address this, we propose HieraVid, a hierarchical pruning framework that progressively and dynamically reduces visual redundancy. Based on two observations that videos possess the segment-frame structure and LLMs internally propagate multi-modal information unidirectionally, we decompose pruning into three levels: 1) segment-level, where video tokens are first temporally segmented and spatially merged; 2) frame-level, where similar frames within the same segment are jointly pruned to preserve diversity; 3) layer-level, redundancy gradually shrinks as LLM layer increases w/o compromising performance. We conduct extensive experiments on four widely used video understanding benchmarks to comprehensively evaluate the effectiveness of HieraVid. Remarkably, with only 30% of tokens retained, HieraVid achieves new state-of-the-art performance, while maintaining over 98% and 99% of the performance of LLaVA-Video-7B and LLaVA-OneVision-7B, respectively. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_01881 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | HieraVid: Hierarchical Token Pruning for Fast Video Large Language Models Guo, Yansong Zhu, Chaoyang Ji, Jiayi Lin, Jianghang Cao, Liujuan Computer Vision and Pattern Recognition Computation and Language Video Large Language Models (VideoLLMs) have demonstrated impressive capabilities in video understanding, yet the massive number of input video tokens incurs a significant computational burden for deployment. Existing methods mainly prune video tokens at input level while neglecting the inherent information structure embedded in videos and large language models (LLMs). To address this, we propose HieraVid, a hierarchical pruning framework that progressively and dynamically reduces visual redundancy. Based on two observations that videos possess the segment-frame structure and LLMs internally propagate multi-modal information unidirectionally, we decompose pruning into three levels: 1) segment-level, where video tokens are first temporally segmented and spatially merged; 2) frame-level, where similar frames within the same segment are jointly pruned to preserve diversity; 3) layer-level, redundancy gradually shrinks as LLM layer increases w/o compromising performance. We conduct extensive experiments on four widely used video understanding benchmarks to comprehensively evaluate the effectiveness of HieraVid. Remarkably, with only 30% of tokens retained, HieraVid achieves new state-of-the-art performance, while maintaining over 98% and 99% of the performance of LLaVA-Video-7B and LLaVA-OneVision-7B, respectively. |
| title | HieraVid: Hierarchical Token Pruning for Fast Video Large Language Models |
| topic | Computer Vision and Pattern Recognition Computation and Language |
| url | https://arxiv.org/abs/2604.01881 |