Exploring Fairness across Fine-Grained Attributes in Large Vision-Language Models
This work addresses fairness concerns for users and developers of LVLMs, identifying biases in under-explored attributes, though it is incremental as it builds on existing fairness studies.
The study tackled the problem of fairness in Large Vision-Language Models (LVLMs) by evaluating biases across fine-grained attributes beyond traditional demographics, revealing that LVLMs exhibit biased outputs and that cultural, environmental, and behavioral factors have a more pronounced impact on decision-making.
The rapid expansion of applications using Large Vision-Language Models (LVLMs), such as GPT-4o, has raised significant concerns about their fairness. While existing studies primarily focus on demographic attributes such as race and gender, fairness across a broader range of attributes remains largely unexplored. In this study, we construct an open-set knowledge base of bias attributes leveraging Large Language Models (LLMs) and evaluate the fairness of LVLMs across finer-grained attributes. Our experimental results reveal that LVLMs exhibit biased outputs across a diverse set of attributes and further demonstrate that cultural, environmental, and behavioral factors have a more pronounced impact on LVLM decision-making than traditional demographic attributes.