X-VLA Soft Cloth Folding Dataset
- Provider
- X-VLA authors (Tsinghua AIR) via Hugging Face LeRobot
- License
- Apache-2.0 (source)
- Size
- 123 files
- Format
- Parquet (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/lerobot/xvla-soft-fold)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Robotics
- Source
- https://arxiv.org/abs/2510.10274
About X-VLA Soft Cloth Folding Dataset
Real-world bimanual cloth-folding demonstrations released with X-VLA, a soft-prompted transformer for cross-embodiment vision-language-action modelling (Zheng et al., arXiv:2510.10274), and hosted in LeRobot format. meta/info.json lists 1,542 episodes and about 2.85 million frames at 20 fps for a single folding task, with a high camera and left and right wrist cameras as MP4 video, plus robot state, actions and timestamps in Parquet. The card notes that left- and right-arm action labels are not perfectly synchronised and recommends re-interpolating them from the provided timestamps following the X-VLA code. Measured 2026-11-14 at about 54 GB. Load with lerobot's LeRobotDataset or read the Parquet with pandas or pyarrow. Useful for deformable-object manipulation, long-horizon bimanual imitation learning and fine-tuning vision-language-action models. The card states the licence as Apache 2.0; cite the X-VLA paper.
How to load X-VLA Soft Cloth Folding Dataset in Python
from datasets import load_dataset
ds = load_dataset("lerobot/xvla-soft-fold")
print(ds)Data structure
File types in sampled files: Parquet
Files
Example file: data/chunk-000/file-000.parquet
Analyze this dataset without downloading it
On Scigantic, X-VLA Soft Cloth Folding Dataset opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Robotics datasets
- PhysicalAI Robotics Manipulation SingleArm
- DexCap: Hand Motion Capture Data for Dexterous Manipulation
- NuRec AV Object Benchmark
- Connected Eco-Driving, Fort Collins 2019: Synchronized Vehicle, ADAS, V2V and Traffic-Signal (V2I) Data for Velocity Prediction (DOE EE0008468)
- PhysicalAI Robotics Mindmap Franka Mug In Drawer
- Arena GR 1 Manipulation Task V 3
- CANOE Dataset
- Digit Pose Estimation