Shawn/Yuxuan Tong (童雨轩)

👋 Hi! I am a senior undergraduate at Department of Computer Science and Technology (DCST), Tsinghua University (THU), and a core contributor of verl, a large-scale reinforcement learning library for LLMs. I used to lead the infrastructure and contribute to the algorithm of DAPO: an Open-Source LLM Reinforcement Learning System at Scale at Seed-Infrastructures, ByteDance, advised by Chi Zhang and Haibin Lin lead the project Demystifying Long Chain-of-Thought Reasoning in LLMs at LTI, CMU, advised by Prof. Graham Neubig and Dr. Xiang Yue. lead DART-Math at HKUST-NLP, advised by Prof. Junxian He ... E-mail: tongyuxuan361@gmail.com / tongyx21@mails.tsinghua.edu.cn [Curriculum Vitae (CV)] [GitHub] [Google Scholar] [X(Twitter)] [Blog] [Notion]

Research Interests

I aim to build AI systems capable of independently conducting or substantially assisting humans in complex reasoning in long context.

Specifically, I am interested in the following topics:

• Large Language Model (LLM)

• AI for (Complex) Reasoning like (Advanced)

  • Education (e.g. Eureka Labs)

  • Research (e.g. SciCode-Bench)

  • Software Engineering (e.g. SWE-Bench)

• Reinforcement Learning for Large Language Models (e.g. OpenAI o1)

• Evaluation & Alignment（e.g. Scalable Oversight)

• Hardware-Aware Algorithm Design (e.g. Flash Attention)

Publications

(*) denotes co-first authors

DAPO: an Open-Source LLM Reinforcement Learning System at Scale

Qiying Yu*, Zheng Zhang*, Yuxuan Tong*\(^{\dagger}\), Guangming Sheng*\(^{\dagger}\), …, Yonghui Wu, Mingxuan Wang

Preprint, under review (\(\dagger\) denotes infrastructure lead)

[🏠 Homepage] [📝 Paper@arXiv] [🐱 Code@GitHub] [🤗 Datasets&Models@HF]

Demystifying Long Chain-of-Thought Reasoning in LLMs

Edward Yeo*, Yuxuan Tong*, Xinyao Niu, Graham Neubig, Xiang Yue

Accepted by ICML 2025; Awarded as Best Paper by ICLR 2025 FM-Wild Workshop

[📝 Paper@arXiv] [🧠 Publication@ICML] [🐱 Code@GitHub] [🤗 Datasets&Models@HF]
[🐦 Thread@X(Twitter)] [📑 BibTeX]

DART-Math: Difficulty-Aware Rejection Tuning for Mathematical Problem-Solving

Yuxuan Tong, Xiwen Zhang, Rui Wang, Ruidong Wu, Junxian He

Accepted by NeurIPS 2024

[📝 Paper@arXiv] [🧠 Publication@NeurIPS] [🐱 Code@GitHub] [🤗 Datasets&Models@HF]
[🐦 Thread@X(Twitter)] [🐶 中文博客@知乎] [📊 Leaderboard@PapersWithCode] [📑 BibTeX]

ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation

Jiazheng Xu*, Xiao Liu*, Yuchen Wu, Yuxuan Tong, Qinkai Li, Ming Ding, Jie Tang, Yuxiao Dong

Accepted by NeurIPS 2023

[📝 Paper@arXiv] [🧠 Publication@NeurIPS] [🐱 Code@GitHub] [🖼️ Dataset@HF] [🤖 Model@HF]
[🐦 Thread@X(Twitter)] [🐶 中文博客@知乎] [📑 BibTeX]

Education

• Undergraduate (2021.09 - 2025.07, Expected)

  • Department of Computer Science and Technology (DCST), Tsinghua University (THU)

  • Bachelor's degree in progress

  • GPA: 3.7/4.0

Research Experience

• 2025.02 - Present

  • Seed-Infrastructures, ByteDance

  • Research intern, advised by Chi Zhang and Haibin Lin

  • Working on large-scale reinforcement learning infrastructure.

• 2024.07 - 2025.02

  • Language Technologies Institute (LTI), Carnegie Mellon University (CMU)

  • Research intern, advised by Prof. Graham Neubig and Dr. Xiang Yue.

  • Worked on building and understanding models capable of complex reasoning like OpenAI o1.

• 2023.07 - 2024.06

  • NLP Group, Hong Kong University of Science and Technology (HKUST-NLP)

  • Research intern, advised by Prof. Junxian He.

  • Worked on synthetic data for mathematical reasoning since 2023.12,

  • before which I worked on

    • process supervision, reward modeling, and constrained decoding for mathematical reasoning,

    • model merging.

• 2022.11 - 2023.06

  • Knowledge Engineer Group, Tsinghua University (THUKEG)

  • Research intern, advised by advised by Prof. Jie Tang and Prof. Yuxiao Dong.

  • Worked on reward modeling and RLHF for text-to-image generation.

Projects

• Maintainer and core contributor of the LLM RL library verl (2025.02 - Present)

Awards

• Tsinghua University Research Scholarship (2023, 2024)

• Tsinghua University Comprehensive Merit Scholarship (top 5% undergraduates) (2022)

Academic Services

• Conference Review: NeurIPS 2024, ICLR 2025, ICML 2025, NeurIPS 2025

Skills

• Programming Languages: Python, C/C++, TypeScript/JavaScript, SpinalHDL, System Verilog, etc.

• ML Libraries: PyTorch, DeepSpeed, HuggingFace, vLLM/SGLang, etc.

• Tools: Git, Shell, SLURM, Linux Utilities, LaTeX, etc.

Misc.

• I am from Shengzhou (嵊州), Shaoxing, Zhejiang in China, which is the birthplace of Yue Opera (越剧) and also famous for delicious food like Xiaolongbao (小笼包).

• I spent six years (2009-2015) of my childhood at Yaohua Primary School (耀华小学) in Tianjin, which is quite happy time.

• Before entering Tsinghua University, I studied at Shengzhou High School (嵊州中学), which is located in a small town but full of excellent teachers and classmates.