Feather the Throttle: Revisiting Visual Token Pruning for Vision-Language Model Acceleration.

Mark Endo, Xiaohan Wang, Serena Yeung-Levy

Browse the full ICCV paper archive.