README.md

---
base_model:
- Qwen/Qwen2.5-Coder-7B-Instruct
library_name: transformers
license: mit
metrics:
- accuracy
pipeline_tag: text-generation
---

<div align="center">
  <h1 align="center">
Z1: Efficient Test-time Scaling with Code
</h1>
<p>Train Large Language Model to Reason with Shifted Thinking
</p>
</div>
<p align="center">
  <a href="https://arxiv.org/abs/2504.00810"><b>[📜 Paper]</b></a> •
  <a href="https://huggingface.co/efficientscaling/Z1-7B"><b>[🤗 HF Models]</b></a> •
  <a href="https://github.com/efficientscaling/Z1"><b>[🐱 GitHub]</b></a>
  <!-- <a href="https://9557c5365a6f44dc84.gradio.live"><b>[🐯 Gradio Demo]</b></a> -->
  <br>

  <!-- <a href="#-quick-start">Quick Start</a> • -->
  <!-- <a href="#%EF%B8%8F-citation">Citation</a> -->
</p>

## Model Details
To begin with the shifted thinking mode, please refer to https://github.com/efficientscaling/Z1.


## Evaluation

<p align="left">
    <img src="tts.png" width="700">
    <br>
    <!-- <em>Test-time scaling comparison between Z1-7B and R1-Distill-Qwen-7B. </em> -->
</p>

## Citation
```
@misc{yu2025efficientscaling,
      title={Z1: Efficient Test-time Scaling with Code}, 
      author={Zhaojian Yu and Yinghao Wu and Yilun Zhao and Arman Cohan and Xiao-Ping Zhang},
      year={2025},
      eprint={2504.00810},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2504.00810}, 
}
```