Klear-Reasoner-8B-SFT

✨ Klear-Reasoner-8B-SFT

We present Klear-Reasoner, a model with long reasoning capabilities that demonstrates careful deliberation during problem solving, achieving outstanding performance across multiple benchmarks. We investigate two key issues with current clipping mechanisms in RL: Clipping suppresses critical exploration signals and ignores suboptimal trajectories. To address these challenges, we propose Gradient-Preserving clipping Policy Optimization (GPPO) that gently backpropagates gradients from clipped tokens.

📌 Overview

<div align="center"> <img src="main_result.png" width="100%"/>

_{Benchmark accuracy of Klear-Reasoner-8B on AIME 2024/2025 (avg@64), LiveCodeBench V5 (2024/08/01-2025/02/01, avg@8), and v6 (2025/02/01-2025/05/01, avg@8).}

</div>

Klear-Reasoner is an 8-billion-parameter reasoning model that achieves SOTA performance on challenging math and coding benchmarks:

The model combines:

1. Quality-centric long CoT SFT – distilled from DeepSeek-R1-0528.
2. Gradient-Preserving Clipping Policy Optimization (GPPO) – a novel RL method that keeps gradients from clipped tokens to boost exploration & convergence.

---

Evaluation

When we expand the inference budget to 64K and adopt the YaRN method with a scaling factor of 2.5. Evaluation is coming soon, stay tuned.

📊 Benchmark Results (Pass@1)

We report the average pass@1 results (avg@n), with all other evaluation metrics following the DeepSeek-R1 assessment framework (temperature=0.6, top_p=0.95).

---

🧪 Training

Configure the experimental environment

git clone https://github.com/suu990901/Klear_Reasoner
cd Klear_Reasoner
pip install -r requirements.txt

For the code, we use Firejail for the sandbox environment. Additionally, we implemented multi-process control based on Pebble, enabling automatic resource reclamation upon task timeout. For mathematics, we use math_verify for judging.

Using Ray for Multi-Node Training

For multi-node training​​, ensure ​​all nodes are started and connected via Ray​​ before executing the training script. Below is a brief setup guide for Ray across multiple machines:

Step 1: Start Ray on the Head Node (node0)

On the first node (typically called node0), run:

ray start --head --dashboard-host=0.0.0.0

Get the IP address of the master node.

MASTER_IP=$(hostname -I | awk '{print $1}')

Step 2: Connect Other Nodes (e.g., node1)

On each additional worker node (e.g., node1), run the following, replacing the IP with that of your head node:

ray start --address=\"$MASTER_IP:6379\"

RL Training

Run the following script on the master node to start the training task.

bash recipe/dapo/perf_run_dapo_ours_math.sh # For Math RL
bash recipe/dapo/perf_run_dapo_ours_code.sh # For Code RL

In the startup script, you need to set the following variables:

YOUR_MODEL_PATH="<your_model_path>"
CKPTS_SAVE_DIR="<ckpts_save_path>"
YOUR_TRAIN_FILE="<train_data_path>"
YOUR_TEST_FILE="<test_data_path>"

🤝 Citation

If you find this work helpful, please cite our paper:

@misc{su2025cegppocontrollingentropygradientpreserving,
      title={CE-GPPO: Controlling Entropy via Gradient-Preserving Clipping Policy Optimization in Reinforcement Learning}, 
      author={Zhenpeng Su and Leiyu Pan and Minxuan Lv and Yuntao Li and Wenping Hu and Fuzheng Zhang and Kun Gai and Guorui Zhou},
      year={2025},
      eprint={2509.20712},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2509.20712}, 
}

@article{DBLP:journals/corr/abs-2508-07629,
  author       = {Zhenpeng Su and
                  Leiyu Pan and
                  Xue Bai and
                  Dening Liu and
                  Guanting Dong and
                  Jiaming Huang and
                  Wenping Hu and
                  Fuzheng Zhang and
                  Kun Gai and
                  Guorui Zhou},
  title        = {Klear-Reasoner: Advancing Reasoning Capability via Gradient-Preserving
                  Clipping Policy Optimization},
  journal      = {CoRR},
  volume       = {abs/2508.07629},
  year         = {2025},
  url          = {https://doi.org/10.48550/arXiv.2508.07629},
  doi          = {10.48550/ARXIV.2508.07629},
  eprinttype    = {arXiv},
  eprint       = {2508.07629},
  timestamp    = {Sat, 13 Sep 2025 14:46:27 +0200},
  biburl       = {https://dblp.org/rec/journals/corr/abs-2508-07629.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}