Search papers, labs, and topics across Lattice.
1
0
2
Class-name descriptors can mislead vision-language models, but selecting attributes directly from images boosts accuracy and interpretability significantly.
