Preprint
Aug 2026
E2S-Pruner: Progressive Two-Stage Evidence Fusion for Visual Token Pruning in Vision-Language Models
A spatial novelty constraint is introduced that promotes coverage of distinct image regions and prevents the retained tokens from concentrating in a few locally salient areas and prevents the retained tokens from concentrating in a few locally salient areas in E2S-Pruner.
Taoyu Qian, Qi Wang, D. Shi et al.
· 0 citations