BLOG ILLUSTRATION / EXECUTION STORY

Subtask handoff between two policy levels

A high-level language subtask conditions a separate low-level VLA with current scene and robot state. The two levels receive new observations after action.

Subtask handoff between two policy levelsA high-level VLM interprets intent and feedback. It supplies a language subtask to a separate VLA. The VLA reads current scene and robot state and generates a chunk. Robot action gives new observations to the two levels. This is Hi-Robot’s two-level structure, not the structure of all pi-family models. Subtask down, observation up Intent + feedback Instruction + interaction High-level VLM Interpret / clarify Language subtask Bounded next step Low-level VLA Scene + robot state Action chunk Continuous commands Robot + observe New observations
A high-level VLM interprets the instruction and interaction. It supplies a language subtask to a low-level VLA, which also reads the current scene and robot state. Continuous execution gives new observations for the next high-level decision. The author made this explanatory diagram. The playback is an example. Animation time is not measured runtime. Open full-size SVG