1 paper
Qingxin Xiao, Peilin Zhao, Yangyang Zhao +2
During MLLM decoding, attention often abnormally concentrates on irrelevant image tokens. While existing research dismisses this as invalid noise and forcibly redirects attention t…