SCAPE Framework for Scenario-Conditioned Robot Policy Evaluation

SCAPE uses a simulation-augmented framework to predict how robot-learning policies will perform in specific real-world scenarios. It corrects sim-to-real bias in simulation labels to provide more accurate, scenario-specific performance estimates rather than broad population averages.