Xiaomi-Robotics-0: An Open-Sourced Vision-Language-Action Model with Real-Time Execution

Benchmark Model Rank Results
robot-manipulation-on-calvin-abcd-to-dXiaomi-Robotics-0 (CALVIN ABCD→D)#1Average Length: 4.80
robot-manipulation-on-simplerenv-google-robotXiaomi-Robotics-0 (Visual Matching)#1Success Rate: 85.5
robot-manipulation-on-libero-objectXiaomi-Robotics-0#2Success Rate: 100.0
robot-manipulation-on-simplerenv-widow-xXiaomi-Robotics-0#7Average: 0.792
robot-manipulation-on-liberoXiaomi-Robotics-0#17Success Rate: 98.7
robot-manipulation-on-libero-goalXiaomi-Robotics-0#17Success Rate: 98.8
robot-manipulation-on-libero-longXiaomi-Robotics-0#20Success Rate: 97.2
robot-manipulation-on-libero-spatialXiaomi-Robotics-0#29Success Rate: 98.8
robot-manipulation-on-calvinXiaomi-Robotics-0–avg. sequence length (D to D): 4.75
zero-shot-generalization-on-calvinXiaomi-Robotics-0–Avg. sequence length: 4.75