DIVE: Towards Descriptive and Diverse Visual Commonsense Reasoning
DIVE is a framework that promotes descriptive and diverse visual commonsense reasoning beyond surface-level visual cues.
jun-hyung-park
DIVE is a framework that promotes descriptive and diverse visual commonsense reasoning beyond surface-level visual cues.
We introduce Perceptual-IQ, a benchmark for visual commonsense reasoning that evaluates models on perceptual imagination.