Tango: Taming Visual Signals for Efficient Video Large Language Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914468869439488 |
|---|---|
| author | Yin, Shukang Zhao, Sirui Wang, Hanchao Jia, Baozhi Wang, Xianquan Fu, Chaoyou Chen, Enhong |
| author_facet | Yin, Shukang Zhao, Sirui Wang, Hanchao Jia, Baozhi Wang, Xianquan Fu, Chaoyou Chen, Enhong |
| contents | Token pruning has emerged as a mainstream approach for developing efficient Video Large Language Models (Video LLMs). This work revisits and advances the two predominant token-pruning paradigms: attention-based selection and similarity-based clustering. Our study reveals two critical limitations in existing methods: (1) conventional top-k selection strategies fail to fully account for the attention distribution, which is often spatially multi-modal and long-tailed in magnitude; and (2) direct similarity-based clustering frequently generates fragmented clusters, resulting in distorted representations after pooling. To address these bottlenecks, we propose Tango, a novel framework designed to optimize the utilization of visual signals. Tango integrates a diversity-driven strategy to enhance attention-based token selection, and introduces Spatio-temporal Rotary Position Embedding (ST-RoPE) to preserve geometric structure via locality priors. Comprehensive experiments across various Video LLMs and video understanding benchmarks demonstrate the effectiveness and generalizability of our approach. Notably, when retaining only 10% of the video tokens, Tango preserves 98.9% of the original performance on LLaVA-OV while delivering a 1.88$\times$ inference speedup. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_09547 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Tango: Taming Visual Signals for Efficient Video Large Language Models Yin, Shukang Zhao, Sirui Wang, Hanchao Jia, Baozhi Wang, Xianquan Fu, Chaoyou Chen, Enhong Computer Vision and Pattern Recognition Token pruning has emerged as a mainstream approach for developing efficient Video Large Language Models (Video LLMs). This work revisits and advances the two predominant token-pruning paradigms: attention-based selection and similarity-based clustering. Our study reveals two critical limitations in existing methods: (1) conventional top-k selection strategies fail to fully account for the attention distribution, which is often spatially multi-modal and long-tailed in magnitude; and (2) direct similarity-based clustering frequently generates fragmented clusters, resulting in distorted representations after pooling. To address these bottlenecks, we propose Tango, a novel framework designed to optimize the utilization of visual signals. Tango integrates a diversity-driven strategy to enhance attention-based token selection, and introduces Spatio-temporal Rotary Position Embedding (ST-RoPE) to preserve geometric structure via locality priors. Comprehensive experiments across various Video LLMs and video understanding benchmarks demonstrate the effectiveness and generalizability of our approach. Notably, when retaining only 10% of the video tokens, Tango preserves 98.9% of the original performance on LLaVA-OV while delivering a 1.88$\times$ inference speedup. |
| title | Tango: Taming Visual Signals for Efficient Video Large Language Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.09547 |