ResearchBenchmarksFailure-revealing testing for VLA models on Robotic Manipulation Tasks Simulation (test)Follow90.38Failure DetectionVLAMotor83.349685.17488788.8252May 16, 2026Evaluation ResultsMethodMethodLinksFailure DetectionTrajectory CoverageVLAMotor2026.0590.3882.68VLATest2026.0583.6263.75