dailyai.report

An international research team launched OpenWorldLib to standardize the definition of world models. The framework explicitly excludes text-to-video generators like Sora, arguing these tools lack true physical understanding. This distinction separates superficial visual mimicry from actual environmental simulation. Practitioners can now benchmark models against a rigorous, unified set of criteria.