1 paper
Tinghao Wang, Yichen Guo, Qizhe Zhang +11
Multimodal large language models (MLLMs) have demonstrated impressive performance in multimodal understanding, but processing large numbers of visual tokens results in high computa…