• [2025-08-21] 🎉 Inference Scripts Released! We have released our inference prompts and scripts for embodied pointing abilities.

• [2025-08-20] 🎉 Models and Datasets Released! We have released our pre-trained models, training datasets, and comprehensive evaluation benchmarks. Check out our HuggingFace collection for all available resources.

• [Coming Soon] 📚 Complete training code and detailed training tutorials will be released soon. Stay tuned!

Embodied-R1 is a 3B vision-language model (VLM) designed for general robotic manipulation. Through an innovative "Pointing" mechanism and Reinforced Fine-tuning (RFT) training methodology, it effectively bridges the "seeing-to-doing" gap in robotics, achieving remarkable zero-shot generalization capabilities.

Figure 1: Embodied-R1 framework overview, comprehensive performance evaluation, and zero-shot robotic manipulation demonstrations.

1. Clone the repository:

   git clone https://github.com/pickxiguapi/Embodied-R1.git
   cd Embodied-R1

2. Create and activate Conda environment:

   conda create -n embodied_r1 python=3.11 -y
   conda activate embodied_r1

3. Install dependencies for inference:

   pip install transformers==4.51.3 accelerate
   pip install qwen-vl-utils[decord]

4. Install dependencies for training (optional):

   pip install -r requirements.txt

Run the example code:

cd Embodied-R1/
python inference_example.py

Task instruction: put the red block on top of the yellow block

Before prediction (original image):

After prediction (visualization result):

Task instruction: put pepper in pan

Before prediction (original image):

After prediction (visualization result):

Task instruction: bring me the camel model

Before prediction (original image):

After prediction (visualization result):

Task instruction: loosening stuck bolts

Before prediction (original image):

After prediction (visualization result):

cd eval
python hf_inference_where2place.py
python hf_inference_vabench_point.py
...

We plan to release the complete training code, datasets, and detailed guidelines soon. Stay tuned!

If you use our work in your research, please cite our paper:

@article{yuan2025embodiedr1,
  title={Embodied-R1: Reinforced Embodied Reasoning for General Robotic Manipulation},
  author={Yuan, Yifu and Cui, Haiqin and Huang, Yaoting and Chen, Yibin and Ni, Fei and Dong, Zibin and Li, Pengyi and Zheng, Yan and Hao, Jianye},
  year={2025}
}