Search papers, labs, and topics across Lattice.
1
0
3
6
VLMs are surprisingly bad at visually matching objects unless they can name them, revealing a critical reliance on textual anchors that overshadows their visual processing capabilities.
