Search papers, labs, and topics across Lattice.
Affiliation:
3
0
5
GRA reveals that grounding VLA models in geometric representations from generated videos can outperform traditional methods that attempt to extract control signals from the same data.
