AIM: Adaptive Inference of Multi-Modal LLMs via Token Merging and Pruning.

Yiwu Zhong, Zhuoming Liu, Yin Li, Liwei Wang

Browse the full ICCV paper archive.