BeingBeyond/Being-H05-2B

Being-H0.5: Scaling Human-Centric Robot Learning for Cross-Embodiment Generalization

Being-H0.5 is a foundational VLA model that scales human-centric learning with UniHand-2.0 and a unified action space to enable robust cross-embodiment robot control.

(For our previous Being-H0 version, please visit the being-h0 branch.)

News

• [2026-01-20]: We publish the Being-H0.5! Check our Paper for technical details and Hugging Face Model Collections for pretrained and post-trained models. 🔥🔥🔥
• [2025-08-02]: We release the Being-H0 codebase and pretrained models! Check our Hugging Face Model Collections for more details. 🔥🔥🔥
• [2025-07-21]: We publish Being-H0! Check our paper here. 🌟🌟🌟

Model Checkpoints

Download models from Hugging Face:

Model Type	Model Name	Parameters	Description
VLA Pretrained	Being-H05-2B	2B	Base vision-language-action model (preview)
VLA Specialist	Being-H05-2B_libero	2B	Post-trained on LIBERO benchmark
VLA Specialist	Being-H05-2B_robocasa	2B	Post-trained on RoboCasa kitchen tasks
VLA Generalist	Being-H05-2B_libero_robocasa	2B	Post-trained on both LIBERO and RoboCasa

Note: the vision part is 224px by default.

Setup

Clone repository

git clone https://github.com/BeingBeyond/Being-H05.git
cd Being-H05

Create environment

conda create -n beingh python=3.10
conda activate beingh

Install package

pip install -r requirements.txt
pip install flash-attn --no-build-isolation

Inference

Quick Start

Use the pretrained or post-trained model for robot policy inference:

from BeingH.inference.beingh_policy import BeingHPolicy

# Load model
policy = BeingHPolicy(
    model_path="/path/to/Being-H05-2B_libero",
    device="cuda:0"
)

# Get action from observation
action = policy.predict(
    images=images,
    state=robot_state,
    instruction="task prompt"
)

Inference Server

Start an inference server for real-time robot control:

python BeingH/inference/service.py \
    --model_path /path/to/Being-H05-2B_libero \
    --port 8000 \
    --device cuda:0

Evaluation on Benchmarks

Evaluate the model on LIBERO benchmark:

python BeingH/benchmark/libero/eval_libero.py \
    --model_path /path/to/Being-H05-2B_libero \
    --suite libero_spatial \
    --num_episodes 50

Evaluate the model on RoboCasa benchmark:

python BeingH/benchmark/robocasa/eval_robocasa.py \
    --model_path /path/to/Being-H05-2B_robocasa \
    --task PnPCounterToCab \
    --num_episodes 50

Training

Post-Training on Custom Data

Post-train the pretrained model on your own robot data:

torchrun --nproc_per_node=8 BeingH/train/train.py \
    --mllm_path /path/to/InternVL3_5-2B \
    --expert_path /path/to/Qwen3-0.6B \
    --resume_from /path/to/Being-H05-2B \
    --resume_model_only True \
    --dataset_config_file configs/posttrain/libero/libero_all.yaml \
    --output_dir /path/to/output \
    --max_steps 30000 \
    --save_steps 10000 \
    --learning_rate 1e-4 \
    --action_chunk_length 16

Contributing and Building on Being-H05

We encourage researchers and practitioners to leverage Being-H05 as a foundation for their own experiments and applications. Whether you're adapting Being-H05 to new robotic platforms, exploring novel manipulation tasks, or extending the model to new domains, our modular codebase is designed to support your innovations. We welcome contributions of all kinds - from bug fixes and documentation improvements to new features and model architectures. By building on Being-H05 together, we can advance the field of vision-language-action modeling and enable robots to perform more complex and diverse manipulation tasks. Join us in making robotic manipulation more capable, robust, and accessible to all.

Acknowledgments

Being-H05 builds on the following excellent open-source projects:

• InternVL: Vision-Language model backbone
• Bagel: Training framework
• Qwen: Language model and MoE expert
• LIBERO: Benchmark for lifelong robot learning
• RoboCasa: Large-scale simulation benchmark for everyday tasks

We thank the authors for their contributions to the robotics and machine learning communities.

License

Copyright (c) 2026 BeingBeyond Ltd. and/or its affiliates.

SPDX-License-Identifier: Apache-2.0

Citation

If you find our work useful, please consider citing us and give a star to our repository! 🌟🌟🌟

Being-H05

@article{beingbeyond2026beingh05,
  title={Being-H0.5: Scaling Human-Centric Robot Learning for Cross-Embodiment Generalization}, 
  author={Luo, Hao and Wang, Ye and Zhang, Wanpeng and Zheng, Sipeng and Xi, Ziheng and Xu, Chaoyi and Xu, Haiweng and Yuan, Haoqi and Zhang, Chi and Wang, Yiqing and Feng, Yicheng and Lu, Zongqing},
  journal={arXiv preprint arXiv:2601.12993},
  year={2026}
}

Being-H0

@article{beingbeyond2025beingh0,
  title={Being-h0: vision-language-action pretraining from large-scale human videos},
  author={Luo, Hao and Feng, Yicheng and Zhang, Wanpeng and Zheng, Sipeng and Wang, Ye and Yuan, Haoqi and Liu, Jiazheng and Xu, Chaoyi and Jin, Qin and Lu, Zongqing},
  journal={arXiv preprint arXiv:2507.15597},
  year={2025}
}