Mantis: A Versatile Vision-Language-Action Model with Disentangled Visual Foresight

Benchmark Model Rank Results
robot-manipulation-on-libero-spatialMantis#25Success Rate: 98.8
robot-manipulation-on-libero-objectMantis#46Success Rate: 99.2
robot-manipulation-on-libero-longMantis#59Success Rate: 94.2
robot-manipulation-on-libero-goalMantis#85Success Rate: 94.4
robot-manipulation-on-liberoMantis#91Success Rate: 96.7