Revealing and Reducing Gender Biases in Vision and Language Assistants (VLAs)

Leander Girrbach^1,2,3   Yiran Huang^1,2,3   Stephan Alaniz^1,2,3   Trevor Darrell⁴   Zeynep Akata^1,2,3

¹Technical University of Munich   ² Helmholtz Munich   ³ MCML   ⁴ UC Berkeley

Description

This repository is the official implemention of Revealing and Reducing Gender Biases in Vision and Language Assistants (VLAs).

Abstract

Pre-trained large language models (LLMs) have been reliably integrated with visual input for multimodal tasks. The widespread adoption of instruction-tuned image-to-text vision-language assistants (VLAs) like LLaVA and InternVL necessitates evaluating gender biases. We study gender bias in 22 popular open-source VLAs with respect to personality traits, skills, and occupations. Our results show that VLAs replicate human biases likely present in the data, such as real-world occupational imbalances. Similarly, they tend to attribute more skills and positive personality traits to women than to men, and we see a consistent tendency to associate negative personality traits with men. To eliminate the gender bias in these models, we find that finetuning-based debiasing methods achieve the best tradeoff between debiasing and retaining performance on downstream task. We argue for pre-deploying gender bias assessment in VLAs and motivate further development of debiasing strategies to ensure equitable societal outcomes.

Setup

To run the code, we need to setup (1) the data; (2) the models; and (3) the prompts. We provide example scripts for all 3 steps:

Data Run setup_data.sh. This will download and process all data. Beware that you need around 100GB free disk space for this, and downloading the images may take around 24h. To increase efficiency, modify scripts under ./setup/ accordingly.

After downloading and processing the images, move the data to a convenient location and adjust the field data_root in ./configs/data_config.yaml.

Models Run setup_models.sh $PATH. Replace $PATHwith the location where you store models. This will download all models from hugginface, using git lfs. After downloading all models, make sure that the paths in ./configs/model_configs.yaml are correctly specified.

Prompts Run python make_prompts.py. This will create all necessary prompts to run evaluation. Prerequisite is that data has been correctly set up.

Environment

We recommend using virtual environments, e.g. mamba. Usage:

mamba create -n vla-bias python=3.10 gdown git-lfs;
mamba activate vla-bias
pip install -f requirements.txt

Usage

The basic usage of the benchmark is

python benchmark.py --model $MODEL --prompt-chunk-index $INDEX

Here, $MODELshould be replaced by a valid model and $INDEX by a prompt chunk index. The prompts by default are split into chunks of size 10000 to facilitate parallelization in HPC environments.

Models can be tuned using the scripts tuning.py and prompt_tuning.py. Example usage:

python tuning.py --model $MODEL --lr 0.0001 --threshold 0.05 --num-images-per-dataset 5000 --max-steps 20000

Citation

Please use the following bibtex entry to cite our work:

@inproceedings{girrbach2024revealing,
  title={Revealing and Reducing Gender Biases in Vision and Language Assistants (VLAs)},
  author={Girrbach, Leander and Huang, Yiran and Alaniz, Stephan and Darrell, Trevor and Akata, Zeynep},
  booktitle={arXiv},
  year={2024}
}

Acknowledgements

This work was supported by the ERC (853489 - DEXIM) and Berkeley AI Research (BAIR) Commons. The authors gratefully acknowledge the Gauss Centre for Supercomputing e.V. (www.gausscentre.eu) for funding this project by providing computing time on the GCS Supercomputer JUWELS at Julich Supercomputing Centre (JSC).