--- license: mit task_categories: - robotics pretty_name: RoboTwin 2.0 demonstrations (LeRobot v2.1) tags: - robotwin - lerobot - manipulation - bimanual --- # RoboTwin 2.0 Demonstrations (LeRobot v2.1) Expert demonstrations for the 50 [RoboTwin 2.0](https://github.com/RoboTwin-Platform/RoboTwin) benchmark tasks, generated with the RoboTwin simulator (MIT license) and packaged in LeRobot **v2.1** format. Used to train [FACT](https://huggingface.co/Bariona/fact-wam). Paper: [FACT: Failure-Aware Causal Training for World-Action Models](https://huggingface.co/papers/2608.10232) Project page: [https://fact-wam.github.io](https://fact-wam.github.io) Code: [https://github.com/Bariona/FACT](https://github.com/Bariona/FACT) ## Contents Single archive `robotwin-v2.tar` (84.5 GB, ~169 GB extracted). - sha256: `f3392ca1dae75e7eab937094d8b9d118c622081d70eeed9655f9307b27b61f97` ``` datasets/RoboTwin/ Clean// # demo_clean scenes — 50 tasks × 50 episodes Randomized// # demo_randomized scenes — 50 tasks × 50 episodes ``` Each `/` is a LeRobot v2.1 dataset directory (`meta/`, `data/`, `videos/`). ## Schema - Robot: bimanual Aloha (`robot_type: aloha`), 14-DOF state/action (6 joints + gripper per arm), 15 fps - Features: `observation.state`, `action`, `observation.images.cam_high` (640×480), `observation.images.cam_left_wrist`, `observation.images.cam_right_wrist`, plus LeRobot index/timestamp fields - Language instruction per episode in `meta/` ## Usage ```bash huggingface-cli download Bariona/robotwin-v2 robotwin-v2.tar \ --repo-type dataset --local-dir ./datasets tar -xf ./datasets/robotwin-v2.tar -C ./datasets ``` ## Provenance & license Trajectories were collected with the RoboTwin 2.0 expert data pipeline ([RoboTwin-Platform/RoboTwin](https://github.com/RoboTwin-Platform/RoboTwin), MIT License, © 2025 Tianxing Chen) and converted to LeRobot v2.1. Redistributed under MIT in accordance with the upstream license. If you use this data, please cite RoboTwin.