1 paper
Gexin Huang, Haisheng Fu, Yanting Yang +4
Understanding how vision-language models use visual information requires connecting internal representations to model behavior. We present vlmprobe, an open-source library that int…