paper-with-me

Papers

SparseVLM: Visual Token Sparsification for Efficient Vision-Language Model Inference

2024-10-06 · Yuan Zhang, Chun-Kai Fan, Junpeng Ma, Wenzhao Zheng, Tao Huang, Kuan Cheng, Denis Gudovskiy, Tomoyuki Okuno, Yohei Nakata, Kurt Keutzer, Shanghang Zhang

In vision-language models (VLMs), visual tokens usually consume a significant amount of computational overhead, despite their sparser information density compared to text tokens. To address this, most existing methods learn a network to prune redundant visual tokens and require additional training data. Differently, we propose an efficient training-free token optimization mechanism dubbed SparseVLM without extra parameters or fine-tuning costs. Concretely, given that visual tokens complement text tokens in VLMs for linguistic reasoning, we select visual-relevant text tokens to rate the significance of vision tokens within the self-attention matrix extracted from the VLMs. Then we progressively prune irrelevant tokens. To maximize sparsity while retaining essential information, we introduce a rank-based strategy to adaptively determine the sparsification ratio for each layer, alongside a token recycling method that compresses pruned tokens into more compact representations. Experimental results show that our SparseVLM improves the efficiency of various VLMs across a range of image and video understanding tasks. In particular, LLaVA equipped with SparseVLM reduces 61% to 67% FLOPs with a compression ratio of 78% while maintaining 93% of the accuracy. Our code is available at https://github.com/Gumpest/SparseVLMs.

📄 PDF Abstract BibTeX arXiv:2410.04417

Code (1)

gumpest/sparsevlms 공식 구현 pytorch

Tasks

Language ModelingLanguage ModellingVideo Understanding

Similar Papers 제목 키워드 기반

VASparse: Towards Efficient Visual Hallucination Mitigation for Large Vision-Language Model via Visual-Aware Sparsification

2025-01-11 · Xianwei Zhuang, Zhihong Zhu, Yuxin Xie, Liming Liang 외

Large Vision-Language Models (LVLMs) may produce outputs that are unfaithful to reality, also known as visual hallucinations (VH), which significantly impedes their real-world usage. To alleviate VH, various decoding str…

HallucinationLanguage ModelingLanguage Modelling

FastMMoE: Accelerating Multimodal Large Language Models through Dynamic Expert Activation and Routing-Aware Token Pruning

2025-11-22 · Guoyang Xia, Yifeng Ding, Fengfa Li, Lei Ren 외 arxiv

Multimodal large language models (MLLMs) have achieved impressive performance, but high-resolution visual inputs result in long sequences of visual tokens and substantial inference latency. Reducing redundant visual toke…

VASparse: Towards Efficient Visual Hallucination Mitigation via Visual-Aware Token Sparsification

2025-01-01 · CVPR 2025 1 · Xianwei Zhuang, Zhihong Zhu, Yuxin Xie, Liming Liang 외

Large Vision-Language Models (LVLMs) may produce outputs that are unfaithful to reality, also known as visual hallucinations (VH), which significantly impedes their real-world usage. To alleviate VH, various decoding…

Hallucination

Dynamic Spatial Sparsification for Efficient Vision Transformers and Convolutional Neural Networks

2022-07-04 · Yongming Rao, Zuyan Liu, Wenliang Zhao, Jie zhou 외

In this paper, we present a new approach for model acceleration by exploiting spatial sparsity in visual data. We observe that the final prediction in vision Transformers is only based on a subset of the most informative…

CogVLA: Cognition-Aligned Vision-Language-Action Model via Instruction-Driven Routing & Sparsification

2025-08-28 · Wei Li, Renshan Zhang, Rui Shao, Jie He 외 arxiv

Recent Vision-Language-Action (VLA) models built on pre-trained Vision-Language Models (VLMs) require extensive post-training, resulting in high computational overhead that limits scalability and deployment.We propose Co…