Policies trained from data instead of hand-written controllers.
An open-source 7B vision-language-action model for robot manipulation.
