UTKFace

anchor

Findings

• IC-498Multimodal foundation models exhibit severe group unfairness, with race and age biases more pronounced than gender bias in text-to-image models while gender bias is stronger in image-to-text models [source]