CoolFace
Modelpublic

youngzhong/SOD-GRPO_teacher-4B

sourceHugging Faceapache-2.0updated 4mo agoView on Hugging Face
2likes59downloads
Model Card

<div align="center">

<h1>SOD-GRPO_teacher-4B</h1>

<p> <a href="https://arxiv.org/abs/2605.07725"> <img src="https://img.shields.io/badge/Paper-arXiv-red?logo=arxiv&logoColor=red" alt="Paper on arXiv"/> </a> <a href="https://github.com/YoungZ365/SOD"> <img src="https://img.shields.io/badge/Code-GitHub-black?logo=github&logoColor=white" alt="Code on GitHub"/> </a> <a href="https://huggingface.co/collections/youngzhong/sod-6a03530369d76913c24a4ffb"> <img src="https://img.shields.io/badge/Collection-SOD-yellow?logo=huggingface" alt="HuggingFace Collection"/> </a> <a href="https://huggingface.co/papers/2605.07725"> <img src="https://img.shields.io/badge/Daily%20Paper-SOD-yellow?logo=huggingface&logoColor=yellow" alt="HuggingFace Daily Paper"/> </a> <a href="https://www.alphaxiv.org/abs/2605.07725"> <img src="https://img.shields.io/badge/alphaXiv-2605.07725-purple?logo=arxiv&logoColor=white" alt="alphaXiv"/> </a> </p>

</div>

About

SOD-GRPO_teacher-4B is a 4B agentic reasoning model trained with GRPO (Group Relative Policy Optimization), serving as the teacher model in the SOD distillation framework.

This model is used to distill smaller student models (SOD-0.6B and SOD-1.7B) via the SOD method, which introduces adaptive step-level weighting to handle cascading error propagation in tool-integrated reasoning.

Model Information

AttributeValue
Base ModelQwen3-4B
Training PipelineCold-Start SFT → GRPO
Parameters4B

Related Models

ModelDescription
SOD-0.6BSOD-distilled 0.6B student
SOD-1.7BSOD-distilled 1.7B student
SOD-GRPO_teacher-4BGRPO-trained 4B teacher model (this model)

Performance

We report average@32 over 5 runs on challenging math, science, and code benchmarks.

MethodAIME 2024AIME 2025GPQA-DiamondLiveCodeBench-v6Average
GRPO (This Model)67.6060.4255.1963.1361.59

Distilled Students

ModelAIME 2024AIME 2025GPQA-DiamondLiveCodeBench-v6Average
SOD-0.6B20.8426.1322.1927.7224.22
SOD-1.7B50.8341.7238.7240.6342.98

Acknowledgement

We sincerely thank the authors of DemyAgent-4B and the paper "Demystifying Reinforcement Learning in Agentic Reasoning" (arXiv:2510.11701) for their contribution.

Citation

bibtex
@article{zhong2026sod,
  title={SOD: Step-wise On-policy Distillation for Small Language Model Agents},
  author={Zhong, Qiyong and Zheng, Mao and Song, Mingyang and Lin, Xin and Sun, Jie and Jiang, Houcheng and Wang, Xiang and Fang, Junfeng},
  journal={arXiv preprint arXiv:2605.07725},
  year={2026}
}