Search papers, labs, and topics across Lattice.
1
0
3
Even when deprived of explicit hypernym examples during training, vision-language models can still generalize taxonomic relationships from images, revealing the surprising power of linguistic priors in cross-modal learning.
