Search papers, labs, and topics across Lattice.
1
0
2
4
Tying intermediate reasoning directly to visual evidence allows VLMs to outperform larger models in spatial reasoning tasks.
