1 paper · 1 filter
Xuan He, Cong Wei, Yuhao Cheng +20
Recent studies suggest that video generation models can exhibit certain forms of zero-shot visual reasoning through generated frames. Yet reliable evaluation remains challenging: b…