SANER: Annotation-free Societal Attribute Neutralizer for Debiasing CLIP

TL;DR

SANER is a novel debiasing method for CLIP that removes societal bias by neutralizing attribute information in text features without needing attribute annotations, outperforming existing techniques.

Contribution

Introduces SANER, a simple and effective debiasing approach that eliminates societal bias in CLIP without using attribute annotations or losing attribute-specific information.

Findings

SANER outperforms existing debiasing methods in experiments.

It effectively removes societal bias while preserving attribute-specific information.

SANER does not require attribute annotations during debiasing.

Abstract

Large-scale vision-language models, such as CLIP, are known to contain societal bias regarding protected attributes (e.g., gender, age). This paper aims to address the problems of societal bias in CLIP. Although previous studies have proposed to debias societal bias through adversarial learning or test-time projecting, our comprehensive study of these works identifies two critical limitations: 1) loss of attribute information when it is explicitly disclosed in the input and 2) use of the attribute annotations during debiasing process. To mitigate societal bias in CLIP and overcome these limitations simultaneously, we introduce a simple-yet-effective debiasing method called SANER (societal attribute neutralizer) that eliminates attribute information from CLIP text features only of attribute-neutral descriptions. Experimental results show that SANER, which does not require attribute annotations and preserves original information for attribute-specific descriptions, demonstrates superior debiasing ability than the existing methods.