LIBERO-PRO contains the bddl and init files for robot task configurations under object, relation, semantic, task, and environment perturbations. The dataset extends the original LIBERO benchmark and supports direct integration with the LIBERO-PRO framework. It was uploaded by zhouxueyang and last updated on 2025-12 03.
Use Cases
- Testing robot policy robustness based on object appearance, color, and scale perturbations.
- Evaluating task performance based on relation perturbations between objects.
- Benchmarking semantic understanding based on perturbations to task goals or instructions.
- Assessing environment generalization based on systematic changes to the simulation scene.
Strengths
- Systematically extends a known benchmark (LIBERO) with perturbations across five defined dimensions.
- Designed for direct integration with a specific framework (LIBERO-PRO), which likely reduces setup effort.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
Provenance
- Source
- zhouxueyang on Hugging Face.
- Collection Method
- Likely generated or modified for the LIBERO-PRO framework.
- Time Range
- null
- Freshness
- Last updated 2025-12-03 03:48:32; freshness should be verified.
- Geography
- null