LIBERO_LeRobot_v3

LIBERO LeRobot v3 Dataset Summary `nvidia/LIBERO_LeRobot_v3` is a LeRobotDataset v3.0 conversion of the LIBERO robot manipulation benchmark. LIBERO is designed for studying lifelong robot learning and knowledge transfer across language-conditioned manipulation tasks. This repository packages the LIBERO task suites as LeRobot-compatible datasets with Parquet state/action data, MP4 video observations, and LeRobot metadata. The dataset is organized as five top-level suite folders: `libero_spatial` `libero_object` `libero_goal` `libero_90` `libero_10` Each suite is stored independently in LeRobotDataset v3.0 format. Dataset Details Dataset Description LIBERO, short for LIfelong learning BEnchmark on RObot manipulation, is a benchmark for studying lifelong learning in decision-making and robot manipulation. It focuses on knowledge transfer across tasks involving spatial relations, object manipulation, goal changes, and long-horizon behavior. This version converts LIBERO demonstration data to the LeRobotDataset v3.0 standard. LeRobot v3.0 uses file-based Parquet and MP4 shards plus structured metadata, replacing per-episode files with chunked files that can contain multiple episodes. This dataset is ready for commercial or non-commercial uses. Dataset Owner(s) NVIDIA Corporation Dataset Creation Date 2026-05-26 Dataset Sources Repository: https://huggingface.co/datasets/nvidia/LIBERO_LeRobot_v3 LIBERO project: https://libero-project.github.io/ LIBERO paper: https://arxiv.org/abs/2306.03310 LIBERO GitHub: https://github.com/Lifelong-Robot-Learning/LIBERO LeRobotDataset v3.0 documentation: https://huggingface.co/docs/lerobot/lerobot-dataset-v3 Version: v1.0 License/Terms of Use This dataset is released under the OpenMDW1.1 Intended Usage This dataset is intended for research and development in: imitation learning robot policy learning vision-language-action model training lifelong robot learning knowledge transfer evaluation language-conditioned manipulation LeRobot-compati