Search papers, labs, and topics across Lattice.
Radboud University
1
0
2
LVLMs leak visual text style into semantic inference, meaning the font of a word can change the attributes a model associates with the concept it represents.
