AI Trends

Hao AI Lab Releases ORA v0.1, a Zero-Shot Robot Operation Framework

UCSD Hao AI LabUpdated Nov 19 09:05

AI summary

Hao AI Lab has released ORA v0.1, an open framework for zero-shot robotic manipulation. Paired with DeepSeek V4.1-Flash, it increased the RoboDojo completion rate from 5.2% to 39.9% and the RoboTwin completion rate from 71.5% to 79.6%. ORA enables models to write code and dynamically create tools while using cuRobo for motion planning. The team also released ORA Traces, which includes 7 models, 5 benchmarks, and 3K+ episodes, and launched Project FastRUA.

Why read this

SelectedAI score 65

Multiple sets of completion rates, code patterns, and collision data for the same robotic tasks provide a basis for comparing agent execution methods and improvements in motion planning.

Topics

Tags

Body

Work done by Yuanbo Yang @YuanboYang60742, Junda Chen @Junda_Chen_, Yulun Wu @memset0c, Shaoxiong Duan @noctuafans, Quanquan Peng @QuanquanPeng03, Xiaolong Wang @xiaolonw and Hao Zhang @haozhangml

- Blog https://t.co/fw54TxK4jG - Traces https://t.co/lobGGMN3Mx - FastRUA https://t.co/VqfrX4pxS1 - Code https://t.co/pdMkhNWGK1

Source

UCSD Hao AI Lab