Elamur Research⌕
☰
home / papers / 2025 / rStar-Coder: Scaling Competitive Code Reasoning with a Large-Scale Verified Dataset

rStar-Coder: Scaling Competitive Code Reasoning with a Large-Scale Verified Dataset

Paper · 2025 · Yifei Liu, Li Lyna Zhang, Yi Zhu et al.

Summary

Training LLMs for competitive programming is limited by a lack of hard problems that come with trustworthy input-output tests. Existing human-curated sets have few, simple tests, and synthetic sets mostly target easy function-level tasks. rStar-Coder addresses this with a 418K-problem dataset paired with 580K long-reasoning solutions, all checked against generated tests. The authors cleaned 37.7K expert-written problems from platforms such as Codeforces, IOI and USACO. They then had GPT-4o write new problems, prompting it with each original statement together with its oracle solution so the new problems stay solvable. Test inputs are built by LLM-written generator and validator functions whose size parameters are swept across several orders of magnitude. Outputs are labeled by majority agreement among several QWQ-32B solutions. Qwen models from 1.5B to 14B fine-tuned on this data match or beat much larger reasoning models on LiveCodeBench and USACO 2025. This shows that verified, difficulty-diverse data can partly substitute for model scale.

Key points

  • Dataset: 418,314 verified problems in total. About 37.7K are expert-designed, filtered from 57,215 collected from TACO, APPS, CodeContests, OpenR1 Codeforces, USACO and IOI. About 380K are synthetic. The dataset also includes 580K long-CoT solutions.
  • Test input generation: a three-step process covers input scales from 10^0 to 10^5. Direct GPT-4o prompting never produced inputs larger than 10^3.
  • Output labeling: mutual verification with QWQ-32B labeled outputs correctly 96.8% of the time, against 12.7% for direct GPT-4o input-output generation. This was measured on 3,150 inputs from 64 problems.
  • LiveCodeBench:
    • Qwen2.5-7B improves from 17.4% to 57.3%.
    • Qwen2.5-14B improves from 23.3% to 62.5%, which is 3.1 points above o3-mini (low).
    • The 1.5B model reaches 40.1%.
  • USACO 2025 (average pass@1): rStar-Coder-7B scores 16.15% and rStar-Coder-14B scores 17.19%, both above QWQ-32B (15.62%). In ablations, seed-only and synthetic-only subsets each help, but combining them works best.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🏢MicrosoftRedmond · United States325 works · 142,931 citations🎓Cornell UniversityIthaca · United States120 works · 59,940 citations🏢Microsoft Research AsiaBeijing · China73 works · 60,539 citations🎓University of Science and Technology of ChinaHefei · China108 works · 112,698 citations🎓Dalian University of TechnologyDalian · China12 works · 46 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksInformation Systems · 1,437 worksComputer Science Applications · 102 works

🧭 Topics 1Research subjects and languages the work is about.Reasoning · 1,426 works

Cite

Cite as Yifei Liu, Li Lyna Zhang, Yi Zhu, Bingcheng Dong, Xudong Zhou, Ning Shang, Fan Yang and Mao Yang. 2025. rStar-Coder: Scaling Competitive Code Reasoning with a Large-Scale Verified Dataset. arXiv:2505.21297.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2505.21297 · arXiv 2505.21297

@misc{liu2025rstar,
  title = {{rStar-Coder: Scaling Competitive Code Reasoning with a Large-Scale Verified Dataset}},
  author = {Yifei Liu, Li Lyna Zhang, Yi Zhu, Bingcheng Dong, Xudong Zhou, Ning Shang, Fan Yang and Mao Yang},
  year = {2025},
  doi = {10.48550/arxiv.2505.21297},
  eprint = {2505.21297},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2505.21297}
}

References

33 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Marah Abdin, Sahaj Agarwal, Ahmed Awadallah, Vidhisha Balachandran, Harkirat Behl, Lingjiao Chen, Gustavo de Rosa, Suriya Gunasekar, Mojan Javaheripi, Neel Joshi, et al. Phi-4-reasoning technical report. arXiv preprint arXiv:2504.21318, 2025. find
  2. Wasi Uddin Ahmad, Sean Narenthiran, Somshubra Majumdar, Aleksander Ficek, Siddhartha Jain, Jocelyn Huang, Vahid Noroozi, and Boris Ginsburg. Opencodereasoning: Advancing data distillation for competitive coding. arXiv preprint arXiv:2504.01943, 2025.
  3. Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, and Charles Sutton. Program synthesis with large language models, 2021.
  4. Sahil Chaudhary. Code alpaca: An instruction-following llama model for code generation. GitHub repository, 2023. find
  5. Bei Chen, Fengji Zhang, Anh Nguyen, Daoguang Zan, Zeqi Lin, Jian-Guang Lou, and Weizhu Chen. Codet: Code generation with generated tests. arXiv preprint arXiv:2207.10397, 2022. find
  6. Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, Alex Ray, Raul Puri, Gretchen Krueger, Michael Petrov, Heidy Khlaaf, Girish Sastry, Pamela Mishkin, Brooke Chan, Scott Gray, Nick Ryder, Mikhai… — Evaluating Large Language Models Trained on Code
  7. Tri Dao. FlashAttention-2: Faster attention with better parallelism and work partitioning. 2023.
  8. Xinyu Guan, Li Lyna Zhang, Yifei Liu, Ning Shang, Youran Sun, Yi Zhu, Fan Yang, and Mao Yang. rstar-math: Small llms can master math reasoning with self-evolved deep thinking. arXiv preprint arXiv:2501.04519, 2025.
  9. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948, 2025.
  10. Dan Hendrycks, Steven Basart, Saurav Kadavath, Mantas Mazeika, Akul Arora, Ethan Guo, Collin Burns, Samir Puranik, Horace He, Dawn Song, et al. Measuring coding challenge competence with apps. arXiv preprint arXiv:2105.09938, 2021.
  11. Baizhou Huang, Shuai Lu, Weizhu Chen, Xiaojun Wan, and Nan Duan. Enhancing large language models in coding through multi-perspective self-consistency. arXiv preprint arXiv:2309.17272, 2023. find
  12. Siming Huang, Tianhao Cheng, Jason Klein Liu, Jiaran Hao, Liuyihan Song, Yang Xu, J Yang, JH Liu, Chenchen Zhang, Linzheng Chai, et al. Opencoder: The open cookbook for top-tier code large language models. arXiv preprint arXiv:2411.04905, 2024.
  13. Yiming Huang, Xiao Liu, Yeyun Gong, Zhibin Gou, Yelong Shen, Nan Duan, and Weizhu Chen. Key-point-driven data synthesis with its enhancement on mathematical reasoning. arXiv preprint arXiv:2403.02333, 2024. find
  14. Binyuan Hui, Jian Yang, Zeyu Cui, Jiaxi Yang, Dayiheng Liu, Lei Zhang, Tianyu Liu, Jiajun Zhang, Bowen Yu, Keming Lu, Kai Dang, Yang Fan, Yichang Zhang, An Yang, Rui Men, Fei Huang, Bo Zheng, Yibo Miao, Shanghaoran Quan, Yunlong Feng, Xingzhang Ren, Xuancheng Ren, Jingren Zhou, and Junyang Lin. Qwen… — Qwen2.5-Coder Technical Report
  15. Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Helyar, Aleksander Madry, Alex Beutel, Alex Carney, et al. Openai o1 system card. arXiv preprint arXiv:2412.16720, 2024.
  16. Naman Jain, King Han, Alex Gu, Wen-Ding Li, Fanjia Yan, Tianjun Zhang, Sida Wang, Armando Solar-Lezama, Koushik Sen, and Ion Stoica. Livecodebench: Holistic and contamination free evaluation of large language models for code. arXiv preprint arXiv:2403.07974, 2024.
  17. Jia LI, Edward Beeching, Lewis Tunstall, Ben Lipkin, Roman Soletskyi, Shengyi Costa Huang, Kashif Rasul, Longhui Yu, Albert Jiang, Ziju Shen, Zihan Qin, Bin Dong, Li Zhou, Yann Fleureau, Guillaume Lample, and Stanislas Polu. Numinamath. [https://huggingface.co/AI-MO/NuminaMath-CoT](https://github.co… find
  18. Rongao Li, Jie Fu, Bo-Wen Zhang, Tao Huang, Zhihong Sun, Chen Lyu, Guang Liu, Zhi Jin, and Ge Li. Taco: Topics in algorithmic code generation dataset. arXiv preprint arXiv:2312.14852, 2023. find
  19. Rongao Li, Jie Fu, Bo-Wen Zhang, Tao Huang, Zhihong Sun, Chen Lyu, Guang Liu, Zhi Jin, and Ge Li. Taco: Topics in algorithmic code generation dataset. arXiv preprint arXiv:2312.14852, 2023. find
  20. Yujia Li, David Choi, Junyoung Chung, Nate Kushman, Julian Schrittwieser, Rémi Leblond, Tom Eccles, James Keeling, Felix Gimeno, Agustin Dal Lago, et al. Competition-level code generation with alphacode. Science, 378(6624):1092–1097, 2022.
  21. Jiawei Liu, Chunqiu Steven Xia, Yuyao Wang, and Lingming Zhang. Is your code generated by chatGPT really correct? rigorous evaluation of large language models for code generation. In Thirty-seventh Conference on Neural Information Processing Systems, 2023.
  22. Ziyang Luo, Can Xu, Pu Zhao, Qingfeng Sun, Xiubo Geng, Wenxiang Hu, Chongyang Tao, Jing Ma, Qingwei Lin, and Daxin Jiang. Wizardcoder: Empowering code large language models with evol-instruct. In The Twelfth International Conference on Learning Representations, 2024. find
  23. Roy Huang Ameen Patel Alpay Ariyak Qingyang Wu Xiaoxiang Shi Rachel Xin Colin Cai Maurice Weber Ce Zhang Li Erran Li Raluca Ada Popa Ion Stoica Michael Luo, Sijun Tan. Deepcoder: A fully open-source 14b coder at o3-mini level. https://pretty-radio-b75.notion.site/ DeepCoder-A-Fully-Open-Source-14B-C… ↗
  24. Guilherme Penedo, Lewis Tunstall, Anton Lozhkov, Hynek Kydlicek, Edward Beeching, Loubna Ben Allal, Quentin Gallouedec, Leandro von Werra, Agustín Piqueres Lajarín, and Nathan Habib. Open r1: Update 3, 2025. find
  25. Quan Shi, Michael Tang, Karthik Narasimhan, and Shunyu Yao. Can language models solve olympiad programming? arXiv preprint arXiv:2404.10952, 2024.
  26. OpenThoughts Team. Open Thoughts. https://open-thoughts.ai, January 2025. ↗
  27. Qwen Team. Qwq-32b: Embracing the power of reinforcement learning, March 2025. find
  28. Yuxiang Wei, Zhe Wang, Jiawei Liu, Yifeng Ding, and Lingming Zhang. Magicoder: Empowering code generation with OSS-instruct. In Proceedings of the 41st International Conference on Machine Learning, volume 235 of Proceedings of Machine Learning Research, pages 52632–52657. PMLR, 21–27 Jul 2024. find
  29. Zhangchen Xu, Yang Liu, Yueqin Yin, Mingyuan Zhou, and Radha Poovendran. Kodcode: A diverse, challenging, and verifiable synthetic dataset for coding, 2025. find
  30. Zhaojian Yu, Xin Zhang, Ning Shang, Yangyu Huang, Can Xu, Yishujie Zhao, Wenxiang Hu, and Qiufeng Yin. Wavecoder: Widespread and versatile enhancement for code large language models by instruction tuning. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Vol… find
  31. Huaye Zeng, Dongfu Jiang, Haozhe Wang, Ping Nie, Xiaotong Chen, and Wenhu Chen. Acecoder: Acing coder rl via automated test-case synthesis. arXiv preprint arXiv:2502.01718, 2025. find
  32. Kexun Zhang, Danqing Wang, Jingtao Xia, William Yang Wang, and Lei Li. Algo: Synthesizing algorithmic programs with generated oracle verifiers. Advances in Neural Information Processing Systems, 36:54769–54784, 2023. find
  33. Xiaojiang Zhang, Jinghui Wang, Zifei Cheng, Wenhao Zhuang, Zheng Lin, Minglei Zhang, Shaojie Wang, Yinghan Cui, Chao Wang, Junyi Peng, et al. Srpo: A cross-domain implementation of large-scale reinforcement learning on llm. arXiv preprint arXiv:2504.14286, 2025. *Table 6: Ablation on scaling dimensi… — [arXiv:2504.14286](https://arxiv.org/abs/2504.14286) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/liu-2025-rstar-coder-scaling-competitive/