SOURCE-LINKED INTELLIGENCE

Centering before Pruning: Lightweight Geometry Correction for Diversity-Based Visual Token Pruning in LVLMs

arXiv · AI, language, vision and robotics · article · Aug 31, 2026 · UTC

Large vision-language models (LVLMs) incur substantial inference costs due to their long and highly redundant visual-token sequences. Diversity-based pruning mitigates this cost by selecting token subsets based on pairwise cosine similarity. We find, however, that similarities between raw visual tokens are strongly concentrated in the positive range, limiting their ability to distinguish non-redundant tokens. A natural way to improve this resolution is to center token features before computing cosine similarity. Centering indeed reveals a substantially richer pairwise structure, yet unexpected

Read original source ↗ Open in workspace

recordType

paper

region

Global