CaP-X open-sourced: large language model agents enter the physical world
Original title:The power of the Claw, in the palm of a robot hand. Agentic robotics is here! Today, we open-source ...
AI summary
CaP-X is an open-source embodied intelligence system that enables large language model agents to enter the physical world through robotic arms and humanoid robots. The system integrates perception APIs such as SAM3 and Molmo with control interfaces including IK solvers and grasp planning, and can automatically synthesize skill libraries. The research releases the CaP-Gym benchmark (187 manipulation tasks) and CaP-Bench (evaluating 12 frontier models), and proposes the zero-shot framework CaP-Agent0 and the reinforcement learning approach CaP-RL, which raised a 7B model's success rate from 20% to 72% in only 50 iterations. The technology was launched by the team that previously developed the Minecraft agent Voyager.
Key Takeaways
- CaP-X is an open-source embodied intelligence system that enables large language model agents to…
- The system integrates perception APIs such as SAM3 and Molmo with control interfaces including IK…
- The research releases the CaP-Gym benchmark (187 manipulation tasks) and CaP-Bench (evaluating 12…
- The technology was launched by the team that previously developed the Minecraft agent Voyager.
Quick Facts
- Published
- Updated
- Source
- X:Jim Fan (@DrJimFan)
- Category
- Other
NVIDIA Jim Fan 开源 CaP-X,让 Vibe Agent 真正进入物理世界操作机器人
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: CaP-X open-sourced: large language model agents enter the physical world. April 1, 2026. https://new.nat6.net/en/news/cmnw1ysux00zrslc36zorasfw/