Weijie Su

We systematically investigate the design space and scaling property of native Multimodal Large Language Models and introduce a novel MLLM that achieves competitive performance against existing MLLMs.