1 paper
Yifeng Yang, Jubo Feng, Jing Xu +3
Vision-language models enable OOD detection by comparing image alignment with ID labels and negative semantics. Existing negative-label-based methods mainly rely on static negative…