Evaluation benchmarks

Model	Benchmark or workflow	Documentation
LingBot-VLA	Open-loop evaluation	Evaluation
LingBot-VLA	RoboTwin 2.0	Evaluation
LingBot-VLA	LIBERO	Evaluation
LingBot-VLA	GM-100	Evaluation
LingBot-VA	RoboTwin 2.0 and LIBERO	Inference and evaluation
LingBot-Map	Trajectory, pose AUC, depth, and point-cloud evaluation	3D reconstruction benchmark

Benchmark results depend on the dataset, checkpoint, configuration, and evaluation protocol. Use the configuration provided for the corresponding model version when comparing results.