$\texttt{ModSCAN}$: Measuring Stereotypical Bias in Large Vision-Language Models from Vision and Language Modalities

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Jiang, Yukun, Li, Zheng, Shen, Xinyue, Liu, Yugeng, Backes, Michael, Zhang, Yang
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916429465387008
author Jiang, Yukun
Li, Zheng
Shen, Xinyue
Liu, Yugeng
Backes, Michael
Zhang, Yang
author_facet Jiang, Yukun
Li, Zheng
Shen, Xinyue
Liu, Yugeng
Backes, Michael
Zhang, Yang
contents Large vision-language models (LVLMs) have been rapidly developed and widely used in various fields, but the (potential) stereotypical bias in the model is largely unexplored. In this study, we present a pioneering measurement framework, $\texttt{ModSCAN}$, to $\underline{SCAN}$ the stereotypical bias within LVLMs from both vision and language $\underline{Mod}$alities. $\texttt{ModSCAN}$ examines stereotypical biases with respect to two typical stereotypical attributes (gender and race) across three kinds of scenarios: occupations, descriptors, and persona traits. Our findings suggest that 1) the currently popular LVLMs show significant stereotype biases, with CogVLM emerging as the most biased model; 2) these stereotypical biases may stem from the inherent biases in the training dataset and pre-trained models; 3) the utilization of specific prompt prefixes (from both vision and language modalities) performs well in reducing stereotypical biases. We believe our work can serve as the foundation for understanding and addressing stereotypical bias in LVLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2410_06967
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle $\texttt{ModSCAN}$: Measuring Stereotypical Bias in Large Vision-Language Models from Vision and Language Modalities
Jiang, Yukun
Li, Zheng
Shen, Xinyue
Liu, Yugeng
Backes, Michael
Zhang, Yang
Cryptography and Security
Computers and Society
Large vision-language models (LVLMs) have been rapidly developed and widely used in various fields, but the (potential) stereotypical bias in the model is largely unexplored. In this study, we present a pioneering measurement framework, $\texttt{ModSCAN}$, to $\underline{SCAN}$ the stereotypical bias within LVLMs from both vision and language $\underline{Mod}$alities. $\texttt{ModSCAN}$ examines stereotypical biases with respect to two typical stereotypical attributes (gender and race) across three kinds of scenarios: occupations, descriptors, and persona traits. Our findings suggest that 1) the currently popular LVLMs show significant stereotype biases, with CogVLM emerging as the most biased model; 2) these stereotypical biases may stem from the inherent biases in the training dataset and pre-trained models; 3) the utilization of specific prompt prefixes (from both vision and language modalities) performs well in reducing stereotypical biases. We believe our work can serve as the foundation for understanding and addressing stereotypical bias in LVLMs.
title $\texttt{ModSCAN}$: Measuring Stereotypical Bias in Large Vision-Language Models from Vision and Language Modalities
topic Cryptography and Security
Computers and Society
url https://arxiv.org/abs/2410.06967