Update README.md

README.md

@@ -1,3 +1,53 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+datasets:
+- Miaow-Lab/RLVR-Linearity-Dataset
+base_model:
+- deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
+pipeline_tag: text-generation
+---
+
+# Model Card
+
+## 1. Model Details
+This model is the fine-tuned checkpoint described in the paper **"Not All Steps are Informative: On the Linearity of LLMs’ RLVR Training"**. It was trained using Reinforcement Learning (GRPO) to enhance mathematical reasoning capabilities.
+
+- **Paper:** [ArXiv](https://arxiv.org/pdf/2601.04537v1)
+- **Code:** [Github](https://github.com/Miaow-Lab/RLVR-Linearity)
+- **Base Model:** [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B)
+- **Training Method:** GRPO
+
+## 2. Performance
+We evaluated the model on standard math benchmarks. Key results include:
+
+| Benchmark | Avg@64 |
+| :--- | :--- |
+| AIME 2024 | **41.93%** |
+
+
+## 3. Training Details
+
+- **Hyperparameters:**
+  - Learning Rate: `1e-6`
+  - Train Batch Size: `128`
+  - PPO Mini Batch Size: `64`
+  - RL Algorithm: `GRPO`
+- **Compute:** Trained on `32 x H100` GPUs for about `150` hours.
+
+For full training configurations, please refer to the `config.json` or the training scripts in our [GitHub](https://github.com/Miaow-Lab/RLVR-Linearity).
+
+## 4. Citation
+
+If you use this model in your research, please cite our paper:
+
+```bibtex
+@misc{wang2026stepsinformativelinearityllms,
+      title={Not All Steps are Informative: On the Linearity of LLMs' RLVR Training}, 
+      author={Tianle Wang and Zhongyuan Wu and Shenghao Jin and Hao Xu and Wei Chen and Ning Miao},
+      year={2026},
+      eprint={2601.04537},
+      archivePrefix={arXiv},
+      primaryClass={cs.LG},
+      url={https://arxiv.org/abs/2601.04537}, 
+}
+```