Self-Evolving Coding Agents: From Digital Programs to Physical-World Intelligence

Benchmark Model Rank Results
robot-manipulation-on-robocasa365-composite-seenHexaModel v0.1#1Success Rate: 62.3
robot-manipulation-on-robocasa365-composite-unseenHexaModel v0.1#1Success Rate: 39.5
robot-manipulation-on-robocasa365-atomic-seenHexaModel v0.1 (extra data)#2Success Rate: 81.0