Saved in:
Bibliographic Details
Main Authors: Sun, Yizheng, Xin, Yanze, Li, Hao, Sun, Jingyuan, Lin, Chenghua, Batista-Navarro, Riza
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2501.13652
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916647726481408
author Sun, Yizheng
Xin, Yanze
Li, Hao
Sun, Jingyuan
Lin, Chenghua
Batista-Navarro, Riza
author_facet Sun, Yizheng
Xin, Yanze
Li, Hao
Sun, Jingyuan
Lin, Chenghua
Batista-Navarro, Riza
contents Multi-modal Large Language Models (MLLMs) have achieved remarkable success by integrating visual and textual modalities. However, they incur significant computational overhead due to the large number of vision tokens processed, limiting their practicality in resource-constrained environments. We introduce Language-Guided Vision Token Pruning (LVPruning) for MLLMs, an effective yet simple method that significantly reduces the computational burden while preserving model performance. LVPruning employs cross-attention modules to compute the importance of vision tokens based on their interaction with language tokens, determining which to prune. Importantly, LVPruning can be integrated without modifying the original MLLM parameters, which makes LVPruning simple to apply or remove. Our experiments show that LVPruning can effectively reduce up to 90% of vision tokens by the middle layer of LLaVA-1.5, resulting in a 62.1% decrease in inference Tera Floating-Point Operations Per Second (TFLOPs), with an average performance loss of just 0.45% across nine multi-modal benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2501_13652
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LVPruning: An Effective yet Simple Language-Guided Vision Token Pruning Approach for Multi-modal Large Language Models
Sun, Yizheng
Xin, Yanze
Li, Hao
Sun, Jingyuan
Lin, Chenghua
Batista-Navarro, Riza
Computation and Language
Multi-modal Large Language Models (MLLMs) have achieved remarkable success by integrating visual and textual modalities. However, they incur significant computational overhead due to the large number of vision tokens processed, limiting their practicality in resource-constrained environments. We introduce Language-Guided Vision Token Pruning (LVPruning) for MLLMs, an effective yet simple method that significantly reduces the computational burden while preserving model performance. LVPruning employs cross-attention modules to compute the importance of vision tokens based on their interaction with language tokens, determining which to prune. Importantly, LVPruning can be integrated without modifying the original MLLM parameters, which makes LVPruning simple to apply or remove. Our experiments show that LVPruning can effectively reduce up to 90% of vision tokens by the middle layer of LLaVA-1.5, resulting in a 62.1% decrease in inference Tera Floating-Point Operations Per Second (TFLOPs), with an average performance loss of just 0.45% across nine multi-modal benchmarks.
title LVPruning: An Effective yet Simple Language-Guided Vision Token Pruning Approach for Multi-modal Large Language Models
topic Computation and Language
url https://arxiv.org/abs/2501.13652