Salvato in:
Dettagli Bibliografici
Autori principali: Tong, Jintao, Jin, Wenwei, Qin, Pengda, Li, Anqi, Zou, Yixiong, Li, Yuhong, Li, Yuhua, Li, Ruixuan
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2505.19536
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908670149787648
author Tong, Jintao
Jin, Wenwei
Qin, Pengda
Li, Anqi
Zou, Yixiong
Li, Yuhong
Li, Yuhua
Li, Ruixuan
author_facet Tong, Jintao
Jin, Wenwei
Qin, Pengda
Li, Anqi
Zou, Yixiong
Li, Yuhong
Li, Yuhua
Li, Ruixuan
contents Large vision-language models (LVLMs) excel at multimodal understanding but suffer from high computational costs due to redundant vision tokens. Existing pruning methods typically rely on single-layer attention scores to rank and prune redundant visual tokens to solve this inefficiency. However, as the interaction between tokens and layers is complicated, this raises a basic question: Is such a simple single-layer criterion sufficient to identify redundancy? To answer this question, we rethink the emergence of redundant visual tokens from a fundamental perspective: information flow, which models the interaction between tokens and layers by capturing how information moves between tokens across layers. We find (1) the CLS token acts as an information relay, which can simplify the complicated flow analysis; (2) the redundancy emerges progressively and dynamically via layer-wise attention concentration; and (3) relying solely on attention scores from single layers can lead to contradictory redundancy identification. Based on this, we propose FlowCut, an information-flow-aware pruning framework, mitigating the insufficiency of the current criterion for identifying redundant tokens and better aligning with the model's inherent behaviors. Extensive experiments show that FlowCut achieves superior results, outperforming SoTA by 1.6% on LLaVA-1.5-7B with 88.9% token reduction, and by 4.3% on LLaVA-NeXT-7B with 94.4% reduction, delivering 3.2x speed-up in the prefilling stage. Our code is available at https://github.com/TungChintao/FlowCut
format Preprint
id arxiv_https___arxiv_org_abs_2505_19536
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FlowCut: Rethinking Redundancy via Information Flow for Efficient Vision-Language Models
Tong, Jintao
Jin, Wenwei
Qin, Pengda
Li, Anqi
Zou, Yixiong
Li, Yuhong
Li, Yuhua
Li, Ruixuan
Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Large vision-language models (LVLMs) excel at multimodal understanding but suffer from high computational costs due to redundant vision tokens. Existing pruning methods typically rely on single-layer attention scores to rank and prune redundant visual tokens to solve this inefficiency. However, as the interaction between tokens and layers is complicated, this raises a basic question: Is such a simple single-layer criterion sufficient to identify redundancy? To answer this question, we rethink the emergence of redundant visual tokens from a fundamental perspective: information flow, which models the interaction between tokens and layers by capturing how information moves between tokens across layers. We find (1) the CLS token acts as an information relay, which can simplify the complicated flow analysis; (2) the redundancy emerges progressively and dynamically via layer-wise attention concentration; and (3) relying solely on attention scores from single layers can lead to contradictory redundancy identification. Based on this, we propose FlowCut, an information-flow-aware pruning framework, mitigating the insufficiency of the current criterion for identifying redundant tokens and better aligning with the model's inherent behaviors. Extensive experiments show that FlowCut achieves superior results, outperforming SoTA by 1.6% on LLaVA-1.5-7B with 88.9% token reduction, and by 4.3% on LLaVA-NeXT-7B with 94.4% reduction, delivering 3.2x speed-up in the prefilling stage. Our code is available at https://github.com/TungChintao/FlowCut
title FlowCut: Rethinking Redundancy via Information Flow for Efficient Vision-Language Models
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2505.19536