Saved in:
Bibliographic Details
Main Authors: Usama, Muhammad, Asim, Syeda Aishah, Ali, Syed Bilal, Wasim, Syed Talal, Mansoor, Umair Bin
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2504.13690
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908328131559424
author Usama, Muhammad
Asim, Syeda Aishah
Ali, Syed Bilal
Wasim, Syed Talal
Mansoor, Umair Bin
author_facet Usama, Muhammad
Asim, Syeda Aishah
Ali, Syed Bilal
Wasim, Syed Talal
Mansoor, Umair Bin
contents Vision-language models (VLMs) have demonstrated impressive capabilities in understanding and reasoning about visual and textual content. However, their robustness to common image corruptions remains under-explored. In this work, we present the first comprehensive analysis of VLM robustness across 19 corruption types from the ImageNet-C benchmark, spanning four categories: noise, blur, weather, and digital distortions. We introduce two new benchmarks, TextVQA-C and GQA-C, to systematically evaluate how corruptions affect scene text understanding and object-based reasoning, respectively. Our analysis reveals that transformer-based VLMs exhibit distinct vulnerability patterns across tasks: text recognition deteriorates most severely under blur and snow corruptions, while object reasoning shows higher sensitivity to corruptions such as frost and impulse noise. We connect these observations to the frequency-domain characteristics of different corruptions, revealing how transformers' inherent bias toward low-frequency processing explains their differential robustness patterns. Our findings provide valuable insights for developing more corruption-robust vision-language models for real-world applications.
format Preprint
id arxiv_https___arxiv_org_abs_2504_13690
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Analysing the Robustness of Vision-Language-Models to Common Corruptions
Usama, Muhammad
Asim, Syeda Aishah
Ali, Syed Bilal
Wasim, Syed Talal
Mansoor, Umair Bin
Computer Vision and Pattern Recognition
Vision-language models (VLMs) have demonstrated impressive capabilities in understanding and reasoning about visual and textual content. However, their robustness to common image corruptions remains under-explored. In this work, we present the first comprehensive analysis of VLM robustness across 19 corruption types from the ImageNet-C benchmark, spanning four categories: noise, blur, weather, and digital distortions. We introduce two new benchmarks, TextVQA-C and GQA-C, to systematically evaluate how corruptions affect scene text understanding and object-based reasoning, respectively. Our analysis reveals that transformer-based VLMs exhibit distinct vulnerability patterns across tasks: text recognition deteriorates most severely under blur and snow corruptions, while object reasoning shows higher sensitivity to corruptions such as frost and impulse noise. We connect these observations to the frequency-domain characteristics of different corruptions, revealing how transformers' inherent bias toward low-frequency processing explains their differential robustness patterns. Our findings provide valuable insights for developing more corruption-robust vision-language models for real-world applications.
title Analysing the Robustness of Vision-Language-Models to Common Corruptions
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2504.13690