Efficient Process Reward Model Training via Active Learning

The official Implementation for Paper "Efficient Process Reward Model Training via Active Learning".

TL;DR: We achieved SOTA performance on ProcessBench (75.0%) and PRMBench (65.5%) with merely 5% labeling cost compared with Qwen/Qwen2.5-Math-PRM-7B.

@misc{duan2025actprm,
      title={Efficient Process Reward Model Training via Active Learning},
      author={Keyu Duan and Zichen Liu and Xin Mao and Tianyu Pang and Changyu Chen and Qiguang Chen and Michael Qizhe Shieh and Longxu Dou},
      year={2025},
      eprint={2504.10559},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2504.10559},
}