Harvard's Yilun Du presents Unified World Models for reconstructing complete 3D scenes from partial text or video inputs · Digg