Our work on evaluating spatio-temporal reasoning of large vision-language models was accepted to CVPR 2026!