Saved in:
Bibliographic Details
Main Authors: Pang, Yijiang, Hoang, Bao, Zhou, Jiayu
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2403.07888
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909156626137088
author Pang, Yijiang
Hoang, Bao
Zhou, Jiayu
author_facet Pang, Yijiang
Hoang, Bao
Zhou, Jiayu
contents Sub-population shift is a specific type of domain shift that highlights changes in data distribution within specific sub-groups or populations between training and testing. Sub-population shift accounts for a significant source of algorithmic bias and calls for distributional robustness. Recent studies found inherent distributional robustness in multi-modality foundation models, such as the vision-language model CLIP, yet this robustness is vulnerable through parameter fine-tuning. In this paper, we propose leveraging the connection of robustness among different modalities and reshaping the distributional robustness of one modality with another. Specifically, in the context of the distributional robustness of CLIP, we propose to leverage natural language inputs to debias the image feature representations, to improve worst-case performance on sub-populations. Our extensive empirical studies show that image representations debiased by natural language can achieve significant performance improvement and reduction of performance instability under sub-population shifts.
format Preprint
id arxiv_https___arxiv_org_abs_2403_07888
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Cross-modality debiasing: using language to mitigate sub-population shifts in imaging
Pang, Yijiang
Hoang, Bao
Zhou, Jiayu
Computer Vision and Pattern Recognition
Artificial Intelligence
Sub-population shift is a specific type of domain shift that highlights changes in data distribution within specific sub-groups or populations between training and testing. Sub-population shift accounts for a significant source of algorithmic bias and calls for distributional robustness. Recent studies found inherent distributional robustness in multi-modality foundation models, such as the vision-language model CLIP, yet this robustness is vulnerable through parameter fine-tuning. In this paper, we propose leveraging the connection of robustness among different modalities and reshaping the distributional robustness of one modality with another. Specifically, in the context of the distributional robustness of CLIP, we propose to leverage natural language inputs to debias the image feature representations, to improve worst-case performance on sub-populations. Our extensive empirical studies show that image representations debiased by natural language can achieve significant performance improvement and reduction of performance instability under sub-population shifts.
title Cross-modality debiasing: using language to mitigate sub-population shifts in imaging
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2403.07888