Horizon · The guide
Coming soon
Horizon, the second competition, is coming soon: human-video task prompting, in-context adaptation, World Action Models and efficient inference in one competition.
1 min read
Horizon is coming soon
What Horizon is
Watch the task. Model the future. Act. Horizon brings several of Physical AI's newest directions into one competition: human-video task prompting, in-context adaptation, World Action Models, and efficient inference.
A model is shown a human or robot demonstration video and the robot's history, and is asked for the future physical state and the robot's actions.
What it will measure
- Everything in Vector: adaptation, precision, robustness and cross-embodiment.
- Human-video learning: teaching unfamiliar physical work through natural visual demonstrations.
- Real-time World Action Models: keeping frontier capability while pushing inference toward practical closed-loop control.
Start with Vector
Vector is open now: start here, or go straight to what you submit.