🌍 Orca-4B — World Model
BAAI/Orca-4B is a general world foundation model centered on Next-State-Prediction. It learns a unified world latent from video + language and exposes it through modality-specific readouts. This demo runs both its text readout (scene / next-state visual QA) and its image + action readout heads for next-state prediction.
Ask about a scene's current state, the physical dynamics at play, or what is likely to happen next (text readout).
Examples