I am a junior student at School of Electronics Engineering and Computer Science, Peking University. My research interests lie in the development of human-like intelligent agents, with a particular focus on understanding and interpreting their behavior to enhance model control and optimize performance. While my prior work has predominantly been rooted in theoretical research, I am equally passionate about exploring the engineering aspects, aiming to bridge the gap between theoretical advancements and practical implementations to drive more robust, efficient, and human-aligned AI systems.
I am currently looking for 2025 summer internship opportunities!!!
🔥 News
- 2025.03: 🎉 Our paper “When More is Less: Understanding Chain-of-Thought Length in LLMs” is accepted to ICLR 2025 Workshop on Reasoning and Planning for Large Language Models!
- 2024.12: 🍁 I attended NuerIPS 2024 at Vancouver and illustrated our poster.
- 2024.10: 🎉 Our paper “A Theoretical Understanding of Self-Correction through In-context Alignment” has been accepted to NeurIPS 2024!
- 2024.06: 🏆 “A Theoretical Understanding of Self-Correction through In-context Alignment” received the Best Paper Award at ICML Workshop on In-context Learning!
📝 Publications
(*: Equal Contribution)
When More is Less: Understanding Chain-of-Thought Length in LLMs
Yuyang Wu*, Yifei Wang*, Tianqi Du, Stefanie Jegelka, Yisen Wang
- I mathematically model the CoT process as a task decomposition and subtask-solving procedure and demonstrate that a longer CoT is not necessarily better.
- Additionally, I conduct both synthetic and real-world experiments, revealing a U-shaped curve and the scaling behavior of the optimal CoT length.
A Theoretical Understanding of Self-Correction through In-context Alignment
Yifei Wang*, Yuyang Wu*, Zeming Wei, Stefanie Jegelka, Yisen Wang
- Best Paper Award at ICML 2024 Workshop on In-context Learning
- I established the first rigorous understanding of LLMs’ self-correction ability and developed a simple and efficient self-correction algorithm (CaC) that shows significant improvements across different tasks.
🎖 Honors and Awards
- 2024.06 Best Paper Award at ICML 2024 Workshop on In-context Learning
- 2021.12 Silver Medal, Chinese Mathematical Olympiad
📖 Education
- 2022.09 - Present, Peking University, BS in Computer Science
💻 Research Experience
- 2023.10 - Present, Research Intern at ZERO Lab, Peking University
- Researching the in-context abilities in LLMs, including self-correction and chain-of-thought.
- Collaborating with Postdoc Yifei Wang (MIT) and advised by Prof. Yisen Wang (PKU)
💪 Skills
- Programming Languages: Python(proficient), C++(proficient), C#, Core Skills (Git/Linux/TeX/etc.)
- Deep Learning Technologies: Pytorch(proficient), CUDA parallel programming