SOURCE-LINKED INTELLIGENCE

VPRune: Efficient Training-free Pre-LLM Visual Token Pruning

arXiv · AI, language, vision and robotics · article · Sep 21, 2026 · UTC

Visual token pruning is a promising approach to reducing the inference cost of large vision-language models (LVLMs), yet aggressive token reduction often causes substantial performance degradation. We identify three key factors behind this degradation: text-guided selection bias, information loss from discarded tokens, and positional distortion caused by sequence compaction. Based on these observations, we propose \textbf{VPRune}, a training-free pre-LLM pruning framework consisting of visual-only diversity selection, similarity-guided token recycling, and position-preserving restoration. Expe

Read original source ↗ Open in workspace

recordType

paper

region

Global