CogVLA: Cognition-Aligned Vision-Language-Action Model via Instruction-Driven Routing & Sparsification

Benchmark Model Rank Results
robot-manipulation-on-libero-spatialCogVLA#38Success Rate: 98.6
robot-manipulation-on-libero-longCogVLA#43Success Rate: 95.4
robot-manipulation-on-libero-goalCogVLA#63Success Rate: 96.6
robot-manipulation-on-libero-objectCogVLA#63Success Rate: 98.8
robot-manipulation-on-liberoCogVLA#67Success Rate: 97.4