Leveraging Visual Tokens for Extended Text Contexts in Multi-Modal Learning Alex Jinpeng Wang, Min Li