GLINT: Sparse-gated VLM for fine-grained radiology

Arxiv CS.CV (computer vision)

GLINT introduces sparse gating for vision-language alignment in radiology, addressing the scale mismatch between image regions and report findings. The model focuses on fine-grained representations for chest X-rays.

glintvision-language-model

Arxiv CS.CV (computer vision)·Jonggwon ParkSeongeun LeeJunhyun ParkHannah YunHyunwoong KimSohyun JeongHyewon KangByungmu YoonKyoyun Choi

8 days ago