Grasp Any Region: Towards Precise, Contextual Pixel Understanding for Multimodal LLMs Paper • 2510.18876 • Published Oct 21, 2025 • 36