Pre-Conference Talk by ZHOU Kankan | VLStereoSet: A Study of Stereotypical Bias in Pre-trained Vision-Language Models

Recently there has been much interest in adapting foundation models such as ALBERT, RoBERTa, T5, GPT-3 and CLIP for different downstream tasks. These models demonstrate powerful transfer capabilities largely because they have acquired the rich body of knowledge contained in their pre-training data. However, their pre-training data may also contain social biases and stereotypes, especially when the data are crawled from the internet without cleaning. As a result, pre-trained models may “inherit” these biases and stereotypes, affecting the fairness of systems derived from these foundation models for downstream tasks. In this paper we study how to measure stereotypical bias in pre-trained vision-language models and leverage a recently released text-only dataset, StereoSet, which covers a wide range of stereotypical bias, and extend it into a vision-language probing dataset called VLStereoSet to measure stereotypical bias in vision-language models. We analyze the differences between text and image and propose a probing task that detects bias by evaluating a model's tendency to pick stereotypical statements as captions for anti-stereotypical images. We further define several metrics to measure both a vision-language model's overall stereotypical bias and its intra-modal and inter-modal bias.

This is a Pre-Conference talk for the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Conference on Natural Language Processing (AACL 2022).