PyRUA-REAL: Robot Exploration and Adaptive Learning through RSI in the Real World
PyRUA-REAL: Robot Exploration and Adaptive Learning through RSI in the Real World
This research introduces PyRUA-REAL, a framework enabling embodied recursive self-improvement (RSI) by having coding agents retain interaction experience as persistent memory and reusable executable skills. The authors evaluate the system across eight real-world tasks, achieving an average success rate of 91.1%, and on 42 RoboDojo benchmark tasks with an 84.4% average success rate. The work investigates how physical manipulation, execution feedback, and task specifications influence code refinement, robustness, and task transfer, ultimately providing empirical guidance for deploying reusable robot skills without online agent decision-making.
The study empirically evaluates a framework for embodied recursive self-improvement through real-world exploration and code refinement, reporting specific success rates on physical manipulation tasks without independent verification of the underlying coding agents' claims.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported success rate
91.1%
Summed sensor-hours
View original evidence
skills influence robustness and task transfer. We further investigate the efficiency and capability of the resulting programs when deployed without involving an online agent for decision-making. Across eight real-world tasks, PyRUA-REAL achieves an average success rate of 91.1%. In complementary evaluation on 42 tasks from the RoboDojo benchmark, it achieves an average success rate of 84.4%. This study provides empi
Open source S4Reported success rate
84.4%
Summed sensor-hours
View original evidence
skills influence robustness and task transfer. We further investigate the efficiency and capability of the resulting programs when deployed without involving an online agent for decision-making. Across eight real-world tasks, PyRUA-REAL achieves an average success rate of 91.1%. In complementary evaluation on 42 tasks from the RoboDojo benchmark, it achieves an average success rate of 84.4%. This study provides empi
Open source S4number of tasks
8
Reported trials
View original evidence
skills influence robustness and task transfer. We further investigate the efficiency and capability of the resulting programs when deployed without involving an online agent for decision-making. Across eight real-world tasks, PyRUA-REAL achieves an average success rate of 91.1%. In complementary evaluation on 42 tasks from the RoboDojo benchmark, it achieves an average success rate of 84.4%. This study provides empi
Open source S4number of benchmark tasks
42
Reported trials
View original evidence
skills influence robustness and task transfer. We further investigate the efficiency and capability of the resulting programs when deployed without involving an online agent for decision-making. Across eight real-world tasks, PyRUA-REAL achieves an average success rate of 91.1%. In complementary evaluation on 42 tasks from the RoboDojo benchmark, it achieves an average success rate of 84.4%. This study provides empi
Open source S4
- paper: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2611.06768v1 Announce Type: new Abstract: Coding agents powered by models such as GPT-6-Astra connect spatial reasoning with executable robot behavior, raising the question of how such agents can produce lasting improvements in robot capabilities from physical interaction. In this work, we study embodied recursive self-improvement (RSI) through real-world exploration, code refinement, and deployment using PyRUA
Open source S2
skills influence robustness and task transfer. We further investigate the efficiency and capability of the resulting programs when deployed without involving an online agent for decision-making. Across eight real-world tasks, PyRUA-REAL achieves an average success rate of 91.1%. In complementary evaluation on 42 tasks from the RoboDojo benchmark, it achieves an average success rate of 84.4%. This study provides empi
Open source S4
-REAL, a framework in which coding agents retain interaction experience as persistent memory and reusable executable skills that inform subsequent exploration and improvement. Our empirical study places physical manipulation at the center of the improvement process, examining how agents collaborate, what kinds of execution feedback support effective code refinement, and how task specifications and previously acquired
Open source S3
rical guidance for turning physical exploration into reusable robot skills that support continued improvement and robust, efficient deployment.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org