Preprint reports keyframe control gains for Unitree humanoid

A preprint describes KINO, a control framework that uses motion keyframes between vision-language planning and reinforcement-learning control for humanoid loco-manipulation. The authors report that a saliency-based keyframe training method raised end-to-end task success from 44% to 92% when sparse VLM keyframes were used. They evaluated pickup, transport and placement tasks in simulation and on a Unitree G1 humanoid, including one- and two-handed manipulation.