arXiv preprint / cs.CV

DIPrune: Task-Aware Token Pruning with Dual Importance for Efficient Multimodal Language Models

作者：Shuo Yang, Changbai Li, Linlin Yang, Huobin Tan, Rongyu Chen, Tongfei Chen, Tian Wang, Sheng Xu, Baochang Zhang · 单位：Communication University of China, Beihang University, National University of Singapore · 会议/期刊：arXiv preprint · 方向：cs.CV · 发布日期：2026-10-07

2026-10-07视觉语言
新标签打开 AlphaXiv
DIPrune: Task-Aware Token Pruning with Dual Importance for Efficient Multimodal Language Models figure
AlphaXiv 中文概览（可滚动查看）