PyRUA-REAL Lets Dual-Arm Robot Self-Improve Through Code, Hits 91.1% Across 8 Real Tasks

PyRUA-REAL, a framework from Qiang Wang and colleagues, studies embodied recursive self-improvement by having coding agents retain interaction experience as persistent memory and reusable executable skills to inform subsequent exploration. On a dual-arm Franka platform across eight physical tasks the system averages 91.1% success. The paper lands as Anthropic's S-1A warns Claude may 'imminently meet' its own threshold for recursive self-improvement.