AdaReTaKe: Adaptive Redundancy Reduction to Perceive Longer for Video-language Understanding
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909642184982528 |
|---|---|
| author | Wang, Xiao Si, Qingyi Wu, Jianlong Zhu, Shiyu Cao, Li Nie, Liqiang |
| author_facet | Wang, Xiao Si, Qingyi Wu, Jianlong Zhu, Shiyu Cao, Li Nie, Liqiang |
| contents | Multimodal Large Language Models (MLLMs) have revolutionized video understanding, yet are still limited by context length when processing long videos. Recent methods compress videos by leveraging visual redundancy uniformly, yielding promising results. Nevertheless, our quantitative analysis shows that redundancy varies significantly across time and model layers, necessitating a more flexible compression strategy. We propose AdaReTaKe, a training-free method that flexibly reduces visual redundancy by allocating compression ratios among time and layers with theoretical guarantees. Integrated into state-of-the-art MLLMs, AdaReTaKe improves processing capacity from 256 to 2048 frames while preserving critical information. Experiments on VideoMME, MLVU, LongVideoBench, and LVBench datasets demonstrate that AdaReTaKe outperforms existing methods by 2.3% and 2.8% for 7B and 72B models, respectively, with even greater improvements of 5.9% and 6.0% on the longest LVBench. Our code is available at https://github.com/SCZwangxiao/video-FlexReduc.git. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_12559 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | AdaReTaKe: Adaptive Redundancy Reduction to Perceive Longer for Video-language Understanding Wang, Xiao Si, Qingyi Wu, Jianlong Zhu, Shiyu Cao, Li Nie, Liqiang Computer Vision and Pattern Recognition Computation and Language Multimedia Multimodal Large Language Models (MLLMs) have revolutionized video understanding, yet are still limited by context length when processing long videos. Recent methods compress videos by leveraging visual redundancy uniformly, yielding promising results. Nevertheless, our quantitative analysis shows that redundancy varies significantly across time and model layers, necessitating a more flexible compression strategy. We propose AdaReTaKe, a training-free method that flexibly reduces visual redundancy by allocating compression ratios among time and layers with theoretical guarantees. Integrated into state-of-the-art MLLMs, AdaReTaKe improves processing capacity from 256 to 2048 frames while preserving critical information. Experiments on VideoMME, MLVU, LongVideoBench, and LVBench datasets demonstrate that AdaReTaKe outperforms existing methods by 2.3% and 2.8% for 7B and 72B models, respectively, with even greater improvements of 5.9% and 6.0% on the longest LVBench. Our code is available at https://github.com/SCZwangxiao/video-FlexReduc.git. |
| title | AdaReTaKe: Adaptive Redundancy Reduction to Perceive Longer for Video-language Understanding |
| topic | Computer Vision and Pattern Recognition Computation and Language Multimedia |
| url | https://arxiv.org/abs/2503.12559 |