1 paper
Chenggang Rong, Tao Han, Zhiyuan Zhao +5
Counting is a core capability for multimodal large language models (MLLMs), yet there is no unified counting dataset to rigorously evaluate this ability across image, text, and aud…