OxyGen: Unified KV Cache Management for Vision-Language-Action Models under Multi-Task Parallelism

作者：Xiangyu Li, Huaizhi Tang, Xin Ding, Weijun Wang, Ting Cao, Yunxin Liu · 单位：Institute for AI Industry Research (AIR), Tsinghua University, Department of Electronic Engineering, Tsinghua University, University of Science and Technology of China · 会议/期刊：arXiv · 日期：2026-03-15 · 来源：Low-Level Learning-Based Action Modelling / Input Modelling / 2D Vision Language Action Models with Efficiency / KV/Token Cache