BLOG ILLUSTRATION / EXECUTION STORY

Add an optional desired scene

The diagram compares supported visual-subgoal conditioning and a direct mode. The optional desired image is an input with current observations.

Add an optional desired sceneIn the supported visual-goal mode, a lightweight world model generates a target proposal from current scene and task context. The VLA reads the target and current scene. The other trace does not use this optional branch. A target image is not proof of reachability or successful control. Current scene + target Scene + subtask Current observation World model Make a target proposal Subgoal image Desired future scene π0.7 VLA Scene + optional target Action chunk Continuous commands
In one supported π0.7 inference mode, a lightweight world model generates a subgoal proposal from current observation and task context. The VLA conditions continuous action on the current scene and desired target. A visual subgoal is not necessary for all operating modes. The author made this explanatory diagram. The playback is an example, and animation time is not measured runtime. Open full-size SVG