VisualThink-VLA：用于高效低延迟视觉-语言-动作策略的视觉中间推理

Mingjian Gao, Wenqiao Zhang, Yuqian Yuan, Yang Dai, Binhe Yu, Zheqi Lv, Haoyu Zheng, Jiaqi Zhu, Zhiqi Ge, Zixuan Wan, Siliang Tang, Yueting Zhuang

摘要

本文提出VisualThink-VLA框架，通过紧凑的视觉证据接口引导动作预测，避免文本推理的延迟和干扰。采用选择性路由机制学习视觉证据令牌，实现低延迟推理并保持高容量专业化，在多个基准和真实机器人上验证了有效性。

vision-language-actionvisual reasoninglow-latencyembodied controlselective routing

MANIPULATION

📊 2 引用

Yusen Li, Ziwei Wang, Xiangye Zhu 等 12 位作者

Robotics and Computer-Integrated Manufacturing · 2026

MANIPULATION

📊 1 引用

Shaoqing Qin, Lida Zhu, Yanpeng Hao 等 10 位作者

Robotics and Computer-Integrated Manufacturing · 2026

MANIPULATION

📊 0 引用

Bo Li, Yuanbo Zhao, Huijie Xiao 等 6 位作者

Robotics and Computer-Integrated Manufacturing · 2026

MANIPULATION

📊 0 引用

Hao Li, Yuhui Yu, Rui Fu 等 6 位作者

Robotics and Computer-Integrated Manufacturing · 2026