1 paper
Haonan Hu, Shuhao Chen, Weisen Jiang +3
Long video understanding requires VideoLLMs to answer queries about videos spanning tens of minutes to hours, but such long videos produce far more visual tokens than current Video…