Fetching the paper…

MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe · Around