Skip-Vision: Efficient and Scalable Acceleration of Vision-Language Models Via Adaptive Token Skipping.

Weili Zeng, Ziyuan Huang, Kaixiang Ji, Yichao Yan

Browse the full ICCV paper archive.