GM-Skip: Metric-Guided Transformer Block Skipping for Efficient Vision-Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Huang, Lianming, Hu, Haibo, Li, Qiao, He, Xin, Guan, Nan, Xue, Chun Jason
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912553353871360
author Huang, Lianming
Hu, Haibo
Li, Qiao
He, Xin
Guan, Nan
Xue, Chun Jason
author_facet Huang, Lianming
Hu, Haibo
Li, Qiao
He, Xin
Guan, Nan
Xue, Chun Jason
contents Transformer-based Vision-Language Models (VLMs) have achieved impressive performance on tasks such as image captioning, object recognition, and visual reasoning, but their high computational cost hinders deployment in latency-sensitive applications like autonomous driving. We introduce GM-Skip, a flexible and metric-adaptive framework for Transformer block skipping that accelerates VLM inference while preserving output quality. GM-Skip features a greedy, metric-guided block selection strategy that uses metric feedback (e.g., accuracy, CIDEr) to identify redundant layers, along with a reverse-order deletion mechanism that preserves early foundational blocks to avoid performance collapse. To support diverse deployment needs, it incorporates a tunable trade-off between sparsity and performance via a score-sparsity balance objective. Experiments across multiple tasks and datasets, including COCO and CODA, show that GM-Skip consistently improves inference speed while maintaining task performance. On the COCO dataset, GM-Skip improves single-object classification accuracy on the Person category from 19.1 percent to 87.3 percent while skipping more than 40 percent of Transformer blocks. In real-world deployment, it achieves up to 45.4 percent latency reduction on single-object detection when integrated into an autonomous vehicle running Autoware.Universe, validating the effectiveness of its skip configurations and confirming its practical value in accelerating real-world inference.
format Preprint
id arxiv_https___arxiv_org_abs_2508_18227
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle GM-Skip: Metric-Guided Transformer Block Skipping for Efficient Vision-Language Models
Huang, Lianming
Hu, Haibo
Li, Qiao
He, Xin
Guan, Nan
Xue, Chun Jason
Computer Vision and Pattern Recognition
Transformer-based Vision-Language Models (VLMs) have achieved impressive performance on tasks such as image captioning, object recognition, and visual reasoning, but their high computational cost hinders deployment in latency-sensitive applications like autonomous driving. We introduce GM-Skip, a flexible and metric-adaptive framework for Transformer block skipping that accelerates VLM inference while preserving output quality. GM-Skip features a greedy, metric-guided block selection strategy that uses metric feedback (e.g., accuracy, CIDEr) to identify redundant layers, along with a reverse-order deletion mechanism that preserves early foundational blocks to avoid performance collapse. To support diverse deployment needs, it incorporates a tunable trade-off between sparsity and performance via a score-sparsity balance objective. Experiments across multiple tasks and datasets, including COCO and CODA, show that GM-Skip consistently improves inference speed while maintaining task performance. On the COCO dataset, GM-Skip improves single-object classification accuracy on the Person category from 19.1 percent to 87.3 percent while skipping more than 40 percent of Transformer blocks. In real-world deployment, it achieves up to 45.4 percent latency reduction on single-object detection when integrated into an autonomous vehicle running Autoware.Universe, validating the effectiveness of its skip configurations and confirming its practical value in accelerating real-world inference.
title GM-Skip: Metric-Guided Transformer Block Skipping for Efficient Vision-Language Models
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2508.18227