Search papers, labs, and topics across Lattice.
1
0
3
VLMs are surprisingly bad at 3D spatial reasoning in panoramic images, but a new RL-based training method closes the gap.
