PROWBench evaluates whether video models follow program-specified motion and interactions. It contains 170 episodes, 200 camera views, and 600 proxy videos, with prompts, recorded world states, and RGB first frames where available. Each view has three proxy representations: white (coarse 3D), mixed (semantic proxy), and colored-obb (3D boxes). Multi-view episodes have four synchronized cameras. The full Viewer split contains 600 rows, one per camera–representation pair; the other splits are subsets. Full prompts are available through promptfile. previewkind distinguishes RGB first-frame references from proxy previews. Long-video prompts use JSON files or a prompts/ directory of windowed…
Organization
Alaya Lab
AlayaLab
Models in Library0
Datasets in Library2
Models on Hugging Face15
Followers83
Compare the published evidence
This profile connects the 2 model and dataset records shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.
Reported licenses: mit. Open a record to review its specific license and conditions.
Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.
Datasets