PyRUA-REAL Turns Robot Exploration into Reusable Skills
A paper introduces PyRUA-REAL, a framework for embodied recursive self-improvement in which coding agents retain physical interaction experiences as persistent memory and reusable executable skills. The authors report an average success rate of 91.1% across eight real-world tasks and 84.4% across 42 tasks from the RoboDojo benchmark, including deployment without an online agent for decision-making. The paper also examines agent collaboration, execution feedback for code refinement, task specifications, robustness, and skill transfer.
The framework offers an empirical approach to turning real-world robot exploration into reusable skills that can support later code refinement, task transfer, and offline deployment.
Computer Science > Robotics
arXiv:2611.06768 (cs)
[Submitted on 4 Nov 2026]
Title:PyRUA-REAL: Robot Exploration and Adaptive Learning through RSI in the Real World
Authors:Qiang Wang, Jianxin Bi, Shulong Jiang, Ruiyang Si, Wenbo Huang, Kaiqi Chen, Jiaxin Li, Liyun Li, Xuanyi Li, Daquan Zhou
View a PDF of the paper titled PyRUA-REAL: Robot Exploration and Adaptive Learning through RSI in the Real World, by Qiang Wang and 9 other authors