A Glimpse to Compress: Dynamic Visual Token Pruning for Large Vision-Language Models
IEEE Transactions on Circuits and Systems for Video Technology · 2026-10 · 36 (10) · 15539-15552
DOI: 10.1109/tcsvt.2026.3702147 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Circuits and Systems for Video Technology, 2026-10. Verfasst von Quan-Sheng Zeng und 6 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 62 Literaturangaben. Gefördert von: NSFC, Fundamental Research Funds for the Central Universities.
Abstract (Originalsprache)
Visual token compression is critical for Large Vision-Language Models (LVLMs) to efficiently process high-resolution inputs. Existing methods that typically adopt fixed compression ratios cannot adapt to scenes of varying complexity, often causing imprecise pruning that discards informative visual tokens and results in degraded model performance. To address this issue, we introduce a dynamic pruning framework, GlimpsePrune, inspired by human cognition. It takes a data-driven “glimpse” and prunes irrelevant visual tokens in a single forward pass before answer generation. This approach prunes 92.4% of visual tokens while on average nearly preserving the original performance on free-form VQA tasks, causing end-to-end latency and peak memory to reduce to 66% and 67% of the baseline, respectively. By integrating our pruning mechanism into the reinforcement learning fine-tuning stage, the time and memory costs of fine-tuning are also reduced with negligible performance loss. Code is available at https://github.com/HVision-NKU/GlimpsePrune
Förderung
- NSFC
- Fundamental Research Funds for the Central Universities
Zitieren
APA
Zeng, Q., Li, Y., Wang, Q., Jiang, P., Wu, Z., Cheng, M., & Hou, Q. (2026). A Glimpse to Compress: Dynamic Visual Token Pruning for Large Vision-Language Models. IEEE Transactions on Circuits and Systems for Video Technology, 36(10), 15539-15552. https://doi.org/10.1109/tcsvt.2026.3702147
Vancouver
Zeng Q, Li Y, Wang Q, Jiang P, Wu Z, Cheng M, et al. A Glimpse to Compress: Dynamic Visual Token Pruning for Large Vision-Language Models. IEEE Transactions on Circuits and Systems for Video Technology. 2026;36(10):15539-15552. doi:10.1109/tcsvt.2026.3702147
BibTeX
@article{Zeng2026,
author = {Zeng, Quan-Sheng and Li, Yunheng and Wang, Qilong and Jiang, Peng-Tao and Wu, Zuxuan and Cheng, Ming-Ming and Hou, Qibin},
title = {A Glimpse to Compress: Dynamic Visual Token Pruning for Large Vision-Language Models},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
year = {2026},
volume = {36},
number = {10},
pages = {15539--15552},
doi = {10.1109/tcsvt.2026.3702147}
}
EUVIMED