DOGE: Towards Versatile Visual Document Grounding and Referring

Open in new window