●builderRelevant if you're exploring video-to-action pipelines for robotics; the framework exposes which generative models are actually usable as robot policy priors.
●researcherOffers a grounded evaluation methodology for physical reasoning in video models, moving beyond perceptual metrics to execution-based success rates.