Unified Vision-Language-Action Model

Benchmark Model Rank Results
robot-manipulation-on-calvin-abcd-to-dUniVLA (CALVIN ABCD→D)#4Average Length: 4.63
robot-manipulation-on-simplerenv-widow-xUniVLA#20Average: 0.698
robot-manipulation-on-libero-objectUniVLA (LIBERO-Object)#62Success Rate: 98.8
robot-manipulation-on-libero-longUniVLA (LIBERO-Long)#63Success Rate: 94.0
robot-manipulation-on-libero-spatialUniVLA (LIBERO-Spatial)#84Success Rate: 95.4
robot-manipulation-on-libero-goalUniVLA (LIBERO-Goal)#87Success Rate: 93.6
robot-manipulation-on-liberoUniVLA#105Success Rate: 95.5
robot-manipulation-on-calvinUniVLA–avg. sequence length (D to D): 4.41