1 briefs tagged #worldexam.
The benchmark evaluates whether controllable video models infer plausible consequences from scene state.