Action conditioned prediction
Physical AI
Predicting future observations or states while taking proposed actions into account.
In a world model, an action can change the predicted future. Evaluation should examine whether the prediction responds appropriately to that action and preserves relevant physical and temporal relationships. Visually plausible output alone does not establish an accurate account of what the action would cause.
In practice
The predicted object motion should differ when a gripper releases it versus when the gripper keeps holding it.