Latent Noise Mask for Reducing Visual Redundancy in Multimodal Large Language Models figure
AlphaXiv 中文概览(可滚动查看)