1 paper · 1 filter
Guanqi Zhan, Changye Li, Zhijian Liu +4
Visual grounding is an essential capability of Visual Language Models (VLMs) to understand the real physical world. Previous state-of-the-art grounding visual language models usual…