MotorMind：用通用视觉语言模型实现零样本机器人操作
原标题：Paper page - MotorMind: Scaffolding General Vision Language Models for Zero-Shot Robot Manipulation
AI 摘要
该论文提出 MotorMind，一种机器人操作框架，将通用视觉语言模型（VLM）提出的中层动作连接到确定性机器人控制与反馈，并具备异步监控和后台记忆更新。无需任务特定策略训练、编码智能体或 SAM3 等额外 grounding 工具，MotorMind 在 LIBERO-PRO 基础套件上达到 66.7% 成功率，扰动下为 53.8%，而此前零样本方法分别最高仅 13.3% 和 19.2%。在真实 xArm6 机器人上，该接口在直接操作和人类扰动设置下平均成功率达 95%。结果表明，配备合适中层动作表示和异步执行框架的通用 VLM 可有效执行零样本机器人操作。
正文节选
MotorMind: Scaffolding General Vision Language Models for Zero-Shot Robot Manipulation Abstract Vision-language-action (VLA) models have advanced robotic manipulation, but their zero-shot generalization in new tasks and environments remains limited, and their reliance on specialized training keeps them from benefiting directly from rapidly advancing general-purpose vision-language models (VLMs). In parallel, recent agentic robotic systems leverage VLMs for high-level reasoning or coding agents f