Search papers, labs, and topics across Lattice.
4
0
7
3
Qualitative examples show the PhysBrain 1.5 model's ability to produce end-effector trajectories and predict future scenes through spatially aligned RGB, depth, and robot-mask outputs.
Decoupling semantic intent from geometric constraints leads to a 47% reduction in physical hallucinations during human-scene interactions.
Finally, video generation achieves granular control over scene lighting, layout, and camera movement, unlocking new creative possibilities.
