Elamur Research⌕
☰
home / papers / 2024 / O1 Replication Journey — Part 2: Surpassing O1-preview through Simple Distillation, Big Progress or Bitter Lesson?

O1 Replication Journey — Part 2: Surpassing O1-preview through Simple Distillation, Big Progress or Bitter Lesson?

Paper · 2024 · Zhen Huang, Haoyang Zou, Xuefeng Li et al.

Summary

Since OpenAI's O1 appeared, many groups have claimed to replicate its long-form reasoning without saying how they did it. This report from GAIR (Shanghai Jiao Tong University and partners) argues that a simple and rarely acknowledged route explains much of this progress: distilling O1 through its API. The authors prompted O1 to produce extended reasoning traces with reflection and backtracking on difficult math problems. They then used standard supervised fine-tuning to train a base model on several tens of thousands of these traces. The resulting model beat O1-preview on AIME with little engineering effort. Although it saw only math data, it also generalized to open-domain QA, safety and hallucination tests, and it became noticeably less sycophantic. The paper proposes a framework for rating O1 replication efforts by how transparent and reproducible they are. It also warns that relying on distillation limits students to their teacher's ceiling, slows real methodological innovation, and weakens the training of researchers who reason from first principles. The authors call this the "bitter lesson."

Key points

  • The training data comes from Olympiad-level problems drawn from open-source and self-curated sets. Rule-based filters removed problems that need images, problems without labeled answers, and proof questions, leaving only items with numerical answers.
  • GPT-4o-mini rewrote the original solutions into longer, more detailed step-by-step form, following the "reformatted technology" approach of Fan et al., 2024. Distilled O1 long-thought chains then served as SFT targets.
  • The authors compare four ways to synthesize long thoughts: human annotation, multi-agent debate, tree search (the "journey learning" method from Part 1), and distillation. They judge distillation the cheapest reliable option.
  • Because of OpenAI's Terms of Use, the distilled data is for research only and will not be fully released. Related materials are planned at github.com/GAIR-NLP/O1-Journey.

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

🎓New York UniversityNew York · United States233 works · 104,003 citations🎓Shanghai Jiao Tong UniversityShanghai · China232 works · 20,305 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 19,503 worksArtificial Intelligence · 18,805 worksArts and Humanities · 17,959 worksLanguage and Linguistics · 17,789 worksComputer Vision and Pattern Recognition · 3,621 works

Cite

Cite as Zhen Huang, Haoyang Zou, Xuefeng Li, Yixiu Liu, Yuxiang Zheng, Ethan Chern, Shijie Xia, Yiwei Qin, Weizhe Yuan and Pengfei Liu. 2024. O1 Replication Journey — Part 2: Surpassing O1-preview through Simple Distillation, Big Progress or Bitter Lesson?. arXiv:2411.16489.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2411.16489 · arXiv 2411.16489

@misc{huang2024o1,
  title = {{O1 Replication Journey — Part 2: Surpassing O1-preview through Simple Distillation, Big Progress or Bitter Lesson?}},
  author = {Zhen Huang, Haoyang Zou, Xuefeng Li, Yixiu Liu, Yuxiang Zheng, Ethan Chern, Shijie Xia, Yiwei Qin, Weizhe Yuan and Pengfei Liu},
  year = {2024},
  doi = {10.48550/arxiv.2411.16489},
  eprint = {2411.16489},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2411.16489}
}

References

38 references, 22 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, et al. 2021. Program synthesis with large language models. arXiv preprint arXiv:2108.07732.
  2. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter,… — Language Models are Few-Shot Learners
  3. 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
  4. Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde De Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, et al. 2021b. Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374.
  5. I Chern, Steffi Chern, Shiqi Chen, Weizhe Yuan, Kehua Feng, Chunting Zhou, Junxian He, Graham Neubig, Pengfei Liu, et al. 2023. Factool: Factuality detection in generative ai–a tool augmented framework for multi-task and multi-domain scenarios. arXiv preprint arXiv:2307.13528.
  6. Paul F Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. 2017. Deep reinforcement learning from human preferences. Advances in neural information processing systems, 30. find
  7. deepseek. 2024. deepseekr1lite. website. find
  8. Run-Ze Fan, Xuefeng Li, Haoyang Zou, Junlong Li, Shwai He, Ethan Chern, Jiewen Hu, and Pengfei Liu. 2024. Reformatted alignment. arXiv preprint arXiv:2402.12219. find
  9. Luyu Gao, Zhuyun Dai, Panupong Pasupat, Anthony Chen, Arun Tejasvi Chaganty, Yicheng Fan, Vincent Y Zhao, Ni Lao, Hongrae Lee, Da-Cheng Juan, et al. 2022. Rarr: Researching and revising what language models say, using language models. arXiv preprint arXiv:2210.08726.
  10. Suriya Gunasekar, Yi Zhang, Jyoti Aneja, Caio César Teodoro Mendes, Allie Del Giorno, Sivakanth Gopi, Mojan Javaheripi, Piero Kauffmann, Gustavo de Rosa, Olli Saarikivi, et al. 2023. Textbooks are all you need. arXiv preprint arXiv:2306.11644. find
  11. Yancheng He, Shilong Li, Jiaheng Liu, Yingshui Tan, Weixun Wang, Hui Huang, Xingyuan Bu, Hangyu Guo, Chengwei Hu, Boren Zheng, et al. 2024. Chinese simpleqa: A chinese factuality evaluation for large language models. arXiv preprint arXiv:2411.07140.
  12. Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. 2021. Measuring mathematical problem solving with the math dataset. arXiv preprint arXiv:2103.03874.
  13. Geoffrey Hinton. 2015. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531.
  14. Kexin Huang, Xiangyang Liu, Qianyu Guo, Tianxiang Sun, Jiawei Sun, Yaru Wang, Zeyang Zhou, Yixu Wang, Yan Teng, Xipeng Qiu, Yingchun Wang, and Dahua Lin. 2023. Flames: Benchmarking value alignment of chinese large language models. find
  15. kimi. 2024. k0math. website. find
  16. kunlun. 2024. skyworko1. website. find
  17. Junlong Li, Shichao Sun, Weizhe Yuan, Run-Ze Fan, Hai Zhao, and Pengfei Liu. 2023. Generative judge for evaluating alignment. arXiv preprint arXiv:2310.05470.
  18. Hunter Lightman, Vineet Kosaraju, Yuri Burda, Harrison Edwards, Bowen Baker, Teddy Lee, Jan Leike, John Schulman, Ilya Sutskever, and Karl Cobbe. 2024. Let’s verify step by step. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenRevi…
  19. Yixiu Liu, Yuxiang Zheng, Shijie Xia, Jiajun Li, Yi Tu, Chaoling Song, and Pengfei Liu. 2024. Safety-j: Evaluating safety with critique. arXiv preprint arXiv:2407.17075. find
  20. OpenAI. 2024. Learning to reason with llms. find
  21. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training language models to follow instructions with human feedback. Advances in neural information processing systems, 35:27730–27744.
  22. Yiwei Qin, Xuefeng Li, Haoyang Zou, Yixiu Liu, Shijie Xia, Zhen Huang, Yixin Ye, Weizhe Yuan, Hector Liu, Yuanzhi Li, et al. 2024. O1 replication journey: A strategic progress report–part 1. arXiv preprint arXiv:2410.18982.
  23. Charlie Snell, Jaehoon Lee, Kelvin Xu, and Aviral Kumar. 2024. Scaling llm test-time compute optimally can be more effective than scaling model parameters. arXiv preprint arXiv:2408.03314.
  24. Hao Sun, Guangxuan Xu, Jiawen Deng, Jiale Cheng, Chujie Zheng, Hao Zhou, Nanyun Peng, Xiaoyan Zhu, and Minlie Huang. 2022. On the safety of conversational models: Taxonomy, dataset, and benchmark. In Findings of the Association for Computational Linguistics: ACL 2022, Dublin, Ireland, May 22-27, 202…
  25. Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B Hashimoto. 2023. Alpaca: A strong, replicable instruction-following model. Stanford Center for Research on Foundation Models. https://crfm.stanford. edu/2023/03/13/alpaca. html, 3(6):… ↗
  26. LLaMaO1 Team. 2024a. Llamao1. Github. find
  27. OpenO1 Team. 2024b. Openo1. Github. find
  28. B Wang, E Chern, and P Liu. 2023. Chinesefacteval: A factuality benchmark for chinese llms. find
  29. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2022. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171.
  30. Jason Wei, Nguyen Karina, Hyung Won Chung, Yunxin Joy Jiao, Spencer Papay, Amelia Glaese, John Schulman, and William Fedus. 2024. Measuring short-form factuality in large language models. arXiv preprint arXiv:2411.04368.
  31. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824–24837.
  32. Can Xu, Qingfeng Sun, Kai Zheng, Xiubo Geng, Pu Zhao, Jiazhan Feng, Chongyang Tao, and Daxin Jiang. 2023. Wizardlm: Empowering large language models to follow complex instructions. arXiv preprint arXiv:2304.12244.
  33. An Yang, Baosong Yang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Zhou, Chengpeng Li, Chengyuan Li, Dayiheng Liu, Fei Huang, et al. 2024a. Qwen2 technical report. arXiv preprint arXiv:2407.10671. find
  34. An Yang, Beichen Zhang, Binyuan Hui, Bofei Gao, Bowen Yu, Chengpeng Li, Dayiheng Liu, Jianhong Tu, Jingren Zhou, Junyang Lin, et al. 2024b. Qwen2. 5-math technical report: Toward mathematical expert model via self-improvement. arXiv preprint arXiv:2409.12122. — Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvement
  35. Longhui Yu, Weisen Jiang, Han Shi, Jincheng Yu, Zhengying Liu, Yu Zhang, James T Kwok, Zhenguo Li, Adrian Weller, and Weiyang Liu. 2023. Metamath: Bootstrap your own mathematical questions for large language models. arXiv preprint arXiv:2309.12284.
  36. Eric Zelikman, Yuhuai Wu, Jesse Mu, and Noah Goodman. 2022. Star: Bootstrapping reasoning with reasoning. Advances in Neural Information Processing Systems, 35:15476–15488.
  37. Chunting Zhou, Pengfei Liu, Puxin Xu, Srinivasan Iyer, Jiao Sun, Yuning Mao, Xuezhe Ma, Avia Efrat, Ping Yu, Lili Yu, et al. 2024. Lima: Less is more for alignment. Advances in Neural Information Processing Systems, 36. find
  38. Daniel M Ziegler, Nisan Stiennon, Jeffrey Wu, Tom B Brown, Alec Radford, Dario Amodei, Paul Christiano, and Geoffrey Irving. 2019. Fine-tuning language models from human preferences. arXiv preprint arXiv:1909.08593. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/huang-2024-o1-replication-journey-part-2/