StreamingVLA: Streaming Vision-Language-Action Model with Action Flow Matching and Adaptive Early Observation

Benchmark Model Rank Results
robot-manipulation-on-liberoStreamingVLA (AFM)–Success Rate: 97.1
robot-manipulation-on-libero-goalStreamingVLA (AFM)–Success Rate: 98.8
robot-manipulation-on-libero-longStreamingVLA (AFM)–Success Rate: 93.0
robot-manipulation-on-libero-objectStreamingVLA (AFM)–Success Rate: 99.0
robot-manipulation-on-libero-spatialStreamingVLA (AFM)–Success Rate: 97.6