Hi, thank you for the great work!

When training a single image, the prompt is generally designed as:
{'content': '<image>\nThe question.\nThink first, call image_zoom_in_tool if needed, then answer. Format strictly ... ', 'role': 'user'}

How to design prompts when training multiple images together?
Specifically, images in multi_modal_data["image"] are more than one. Does the prompt need to be changed to:
{'content': '<image><image>\nThe question.\nThink first, call image_zoom_in_tool if needed, then answer. Format strictly ... ', 'role': 'user'}
And how to tell the order of images?

Thanks again for your excellent work!