Rethinking Pruning for Vision-Language Models: Strategies for Effective Sparsity
ACM SIGMETRICS Performance Evaluation Review · 26.08.2025 · 53 (2) · 9-14
DOI: 10.1145/3764944.3764948 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM SIGMETRICS Performance Evaluation Review, online erschienen am 27.08.2025. Verfasst von Shwai He und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 53 Literaturangaben.
Abstract (Originalsprache)
Vision-Language Models (VLMs) integrate information from multiple modalities and have shown remarkable success across various tasks. However, deploying large-scale VLMs in resource constrained scenarios is challenging. Pruning followed by finetuning offers a potential solution but remains underexplored for VLMs. This study addresses two key questions: how to distribute sparsity across different modality-specific models, and how to restore the performance of pruned sparse VLMs. Our preliminary studies identified two effective pruning settings: applying the same sparsity to both vision and language models, and pruning only the language models. While LoRA finetuning aims to restore sparse models, it faces challenges due to incompatibility with sparse models, disrupting the pruned sparsity. To overcome these issues, we propose SparseLoRA, which applies sparsity directly to LoRA weights. Our experimental results demonstrate significant improvements, including an 11.3% boost under 2:4 sparsity and a 47.6% enhancement under unstructured 70% sparsity.
Zitieren
APA
He, S., Li, A., & Chen, T. (2025). Rethinking Pruning for Vision-Language Models: Strategies for Effective Sparsity. ACM SIGMETRICS Performance Evaluation Review, 53(2), 9-14. https://doi.org/10.1145/3764944.3764948
Vancouver
He S, Li A, Chen T. Rethinking Pruning for Vision-Language Models: Strategies for Effective Sparsity. ACM SIGMETRICS Performance Evaluation Review. 2025;53(2):9-14. doi:10.1145/3764944.3764948
BibTeX
@article{He2025,
author = {He, Shwai and Li, Ang and Chen, Tianlong},
title = {Rethinking Pruning for Vision-Language Models: Strategies for Effective Sparsity},
journal = {ACM SIGMETRICS Performance Evaluation Review},
year = {2025},
volume = {53},
number = {2},
pages = {9--14},
doi = {10.1145/3764944.3764948}
}
EUVIMED