BBQ-V: Benchmarking Visual Stereotype Bias in Large Multimodal Models

TL;DR

BBQ-V is a comprehensive benchmark designed to evaluate stereotype biases in large multimodal models using real-world images across diverse categories, revealing biases in current models and aiding fairness improvements.

Contribution

Introduces BBQ-V, the most extensive real-image benchmark for assessing stereotype biases in LMMs across multiple categories and sub-categories.

Findings

Top models exhibit significant stereotype biases.

Bias increases in reasoning chains.

Real-world images reveal more nuanced biases.

Abstract

Stereotype biases in Large Multimodal Models (LMMs) perpetuate harmful societal prejudices, undermining the fairness and equity of AI applications. As LMMs grow increasingly influential, addressing and mitigating inherent biases related to stereotypes, harmful generations, and ambiguous assumptions in real-world scenarios has become essential. However, existing datasets evaluating stereotype biases in LMMs often lack diversity, rely on synthetic images, and often have single-actor images, leaving a gap in bias evaluation for real-world visual contexts. To address the gap in bias evaluation using real images, we introduce the BBQ-Vision (BBQ-V), the most comprehensive framework for assessing stereotype biases across nine diverse categories and 50 sub-categories with real and multi-actor images. BBQ-V benchmark contains 14,144 image-question pairs and rigorously evaluates LMMs through carefully curated, visually grounded scenarios, challenging them to reason accurately about visual stereotypes. It offers a robust evaluation framework featuring real-world visual samples, image variations, and open-ended question formats. BBQ-V enables a precise and nuanced assessment of a model's reasoning capabilities across varying levels of difficulty. Through rigorous testing of 19 state-of-the-art open-source (general-purpose and reasoning) and closed-source LMMs, we highlight that these top-performing models are often biased on several social stereotypes, and demonstrate that the thinking models induce more bias in the reasoning chains. This benchmark represents a significant step toward fostering fairness in AI systems and reducing harmful biases, laying the groundwork for more equitable and socially responsible LMMs. Our dataset and evaluation code are publicly available.