Fetching the paper…

Groma: Localized Visual Tokenization for Grounding Multimodal Large Language Models · Around