arXiv / Working Memory / Latent

VisMem: Latent Vision Memory Unlocks Potential of Vision-Language Models

作者：作者信息待提取 · 单位：National University of Singapore, Fudan University, Tsinghua University, Zhejiang University, University of Science and Technology of China · 来源：arXiv · 发布日期：2025/11

工作记忆潜变量记忆
新标签打开 AlphaXiv
VisMem: Latent Vision Memory Unlocks Potential of Vision-Language Models figure
AlphaXiv 中文概览（可滚动查看）