A Woman with a Knife or A Knife with a Woman? Measuring Directional Bias Amplification in Image Captions

ArXi:2503.07878v5 Announce Type: replace-cross When we train models on biased datasets, they not only reproduce data biases, but can worsen them at test time - a phenomenon called bias amplification. Many of the current bias amplification metrics (e.g., BA (MALS), DPA) measure bias amplification only in classification datasets. These metrics are ineffective for image captioning datasets, as they cannot capture the language semantics of a caption. Recent work