Paper summary
Paper summary: VGI-Bench: Probing Visual Intelligence in Video Generation Models
VGI-Bench tests visual reasoning in video generation models with 27 tasks and 810 instances that require a valid process, not just a plausible final frame; the best model, Seedance 2.0, scores 51.0%.
- A video "solution" has to show a valid evolving process, not just end on a plausible final state.
- Later denoising steps mostly refine early guesses rather than correct reasoning errors.