Project Page | Paper | Code
This dataset contains the bddl and init files of LIBERO-PRO configurations under object, relation, semantic, task, and environment perturbations. The dataset supports direct integration with the LIBERO-PRO framework to evaluate Vision-Language-Action (VLA) models beyond rote memorization.
This dataset extends the original LIBERO benchmark by introducing systematic… See the full description on the dataset page:
https://huggingface.co/datasets/zhouxueyang/LIBERO-Pro.