跳到正文
arXiv · cs.AI · Benjamin Pikus, Daniel Saunders, Mircea Mironenco, Salman Mohammadi, Mathew Mathew, Sirou Zhu, Divya Makkar, Xuhua Huang, Qinbo Li, Supreeth Narasimhaswamy, Siyuan Cen, Dejia Xu, Shyamal Buch, Thomas Neff · 5 小时前模型评分49

ZAP:用仿真教机器人动作技能,性能追平 GPT-6 Astra

Pretraining Knows What, Simulation Teaches How: Compositional Generalization for Agentic Robotics

导读

研究者提出零真实数据智能体策略(ZAP,Zero-real Agent Policy),完全在仿真中用简单基础任务和模型自身成功经验训练小型开源视觉语言模型(VLM,Vision-Language Model)补齐动作控制能力,无需前沿教师模型、知识蒸馏或真实机器人数据。

分类置信度

中 · 待人工复核

  • 只有单一媒体来源,需要人工复核。

来源级别:媒体与个人

事件类型:研究发布

证据摘录

  • “We introduce Zero-real Agent Policy (ZAP), a recipe that teaches this missing motion skill, entirely in simulation on simple primitive tasks and from the model's own successes, with no frontier teacher or distillation and no real-world robot data.”

来源:arXiv · cs.AI · arxiv.org