Matteo Pierucci

A new benchmark improves evaluations of models counting objects using language prompts.