Mind-VLA: Instruction-Aware Spatial Representation Alignment for Vision-Language-Action Models

Benchmark Model Rank Results
robot-manipulation-on-calvinMind-VLA–avg. sequence length (D to D): 4.47
robot-manipulation-on-libero-goalMind-VLA–Success Rate: 92.0
robot-manipulation-on-libero-longMind-VLA–Success Rate: 89.5
robot-manipulation-on-libero-objectMind-VLA–Success Rate: 98.0
robot-manipulation-on-libero-spatialMind-VLA–Success Rate: 98.0