Qwen3-30B-A3B-Art

This is the Chain-of-Thought (CoT) efficient version of the Qwen3-30B-A3B-Instruct-2507 model, introduced in the paper The Art of Efficient Reasoning: Data, Reward, and Optimization.

The model is designed to generate short yet accurate reasoning trajectories, reducing computational overhead while maintaining high performance. It was trained on the DeepScaleR-Easy dataset using reward shaping with Reinforcement Learning (RL).

Project Resources

Project Page: https://wutaiqiang.github.io/project/Art
Paper: arXiv:2602.20945

Citation

If you find this work useful, please cite:

@inproceedings{wu2026art,
  title={The Art of Efficient Reasoning: Data, Reward, and Optimization},
  author={Taiqiang Wu and Zenan Xu and Bo Zhou and Ngai Wong},
  year={2026},
  url={https://arxiv.org/pdf/2602.20945}
}