Multi-Stage Vision Token Dropping: Towards Efficient Multimodal Large Language Model

MLLM
VLM
Re-assign community