README.md

---
task_categories:
- robotics
pipeline_tag: robotics
---

<div align="center">
<h1> ALMI-trans </h1>
</div>
<h5 align="center">
    <a href="https://almi-humanoid.github.io/"> 🔥 Homepage</a> | <a href="https://huggingface.co/datasets/TeleEmbodied/ALMI-X">⛁ Dataset</a> | <a href="https://huggingface.co/TeleEmbodied/ALMI-trans">🤗 Models</a> | <a href="https://arxiv.org/abs/2504.14305">📑 Paper</a> | <a href="https://github.com/TeleHuman/ALMI-Open/">💻 Code</a>
</h5>

This repo contains the humanoid foundation model trained using ALMI-X dataset as mentioned in our paper [Adversarial Locomotion and Motion Imitation for Humanoid Policy Learning](https://almi-humanoid.github.io/). We realse the exported models corresponding to `CL-20sl` and `CL-400sl` in our paper.

You can find model training and inference code [here](https://github.com/TeleHuman/ALMI-Open/). Following the guidance you can train and play your own humanoid foundation model in Mujoco simulation with ALMI-X dataset.

# ALMI-X Collection and Model Architecture

![Alt text](assets/almi-x.png)

# Citation

If you find our work helpful, please cite us:

```
        @misc{shi2025almi,
          title={Adversarial Locomotion and Motion Imitation for Humanoid Policy Learning}, 
          author={Jiyuan Shi and Xinzhe Liu and Dewei Wang and Ouyang Lu and Sören Schwertfeger and Fuchun Sun and Chenjia Bai and Xuelong Li},
          year={2025},
          eprint={2504.14305},
          archivePrefix={arXiv},
          primaryClass={cs.RO},
          url={https://arxiv.org/abs/2504.14305}, 
    }
```