rhos-ai/gm100-cobotmagic-lerobot

---
license: mit
tags:
- hdf5
- imitation-learning
---

# Dataset Card for GM-100 Cobot Magic Part (Lerobot2.1 Format)
[](https://www.rhos.ai/research/gm-100)
[](https://arxiv.org/abs/2601.11421)

This is a part of the [Great March 100 (GM-100)](https://www.rhos.ai/research/gm-100) Project.
Raw teleoperation data collected using the Agilex Cobot Magic robot. This dataset is stored in Lerobot2.1 format.

## Notice
This is a preview version of this dataset. A few task prompts may contain minor issues, which we will address in the upcoming full release. We welcome any feedback or suggestions you may have.

## Dataset Description

This repository contains raw robotic manipulation data stored in **Lerobot2.1** format. It is designed for imitation learning tasks using the Agilex Cobot Magic hardware stack.

- **Robot Platform:** Agilex Cobot Magic
- **Data Format:** Lerobot2.1
- **Camera Views:** Top, Left Wrist, Right Wrist

## File Structure

The dataset consists of individual Lerobot2.1 folders, where each folder represents one task.

```text
.
├── task_00001/ # Task 1 Directory
│ ├── meta/ # Metadata for Task 1
│ │ ├── info.json
│ │ ├── tasks.jsonl
│ │ └── ...
│ ├── videos/
│ │ └── chunk-000/
│ │ ├── observation.images.camera_top/
│ │ │ ├── episode_000000.mp4
│ │ │ ├── episode_000001.mp4
│ │ │ └── ...
│ │ ├── observation.images.camera_wrist_left/
│ │ │ ├── episode_000000.mp4
│ │ │ ├── episode_000001.mp4
│ │ │ └── ...
│ │ └── observation.images.camera_wrist_right/
│ │ ├── episode_000000.mp4
│ │ ├── episode_000001.mp4
│ │ └── ...
│ └── data/
│ └── chunk-000/
│ ├── episode_000000.parquet
│ ├── episode_000001.parquet
│ └── ...
├── task_00002/ # Task 2 Directory
│ ├── ...
└── ...
```

## Usage Example
Please refer to the [Lerobot Documentation](https://github.com/huggingface/lerobot) for instructions on how to load and utilize the dataset for imitation learning tasks.

## Citation

```bibtex
@misc{wang2026greatmarch100100,
title={The Great March 100: 100 Detail-oriented Tasks for Evaluating Embodied AI Agents},
author={Ziyu Wang and Chenyuan Liu and Yushun Xiang and Runhao Zhang and Yu Zhang and Qingbo Hao and Hongliang Lu and Houyu Chen and Zhizhong Feng and Kaiyue Zheng and Dehao Ye and Xianchao Zeng and Xinyu Zhou and Boran Wen and Jiaxin Li and Mingyu Zhang and Kecheng Zheng and Qian Zhu and Ran Cheng and Yong-Lu Li},
year={2026},
eprint={2601.11421},
archivePrefix={arXiv},
primaryClass={cs.RO},
url={https://arxiv.org/abs/2601.11421},
}
```