---

Hi there! 👋

I’m Kechi Zhang (张克驰), currently pursuing my Ph.D. at Peking University, with an expected graduation in June 2026.

My research is at the fascinating intersection of AI4SE, LLMs for Code, and Code Generation and Representation through deep learning techniques.

I am passionate about leveraging large language models to enhance software engineering and the way we generate and represent code.

I am now dedicated to building a powerful code agent capable of end-to-end generation of production-ready code, joining outstanding counterparts like Claude Code in striving to transform software engineering development workflows!

Email: zhangkechi@pku.edu.cn

Homepage: Google Scholar

---

🔍 Research Focus

• AI4SE & LLMs for Code
  • Code Generation and Representation through Deep Learning
  • Reinforcement Learning for Long Reasoning Code Models
  • Pre-training, fine-tuning, and alignment of Code LLMs
  • Tool Enhancement and Agent Technology for Code Models
  • Length Extrapolation for Code Models
  • Project-level Code Generation
  • Structural Information-based Code Representation Models

---

📚 Education

• Ph.D. in Computer Software and Theory
  School of Computer Science, Peking University, Beijing, China
  Sept. 2021 - June 2026 (expected)
  Tutor: Prof. Zhi Jin, Prof. Ge Li

• B.S. in Computer Science and Technology
  School of EECS, Peking University, Beijing, China
  Sept. 2017 - July 2021
  GPA: ~3.60 (Top 25%)

---

📝 Selected Publications

• SEAlign | ICSE 2026
  A novel alignment framework aimed at bridging the gap between code generation models and real-world software engineering agent.

• CodeDPO | ACL 2025 Main Conference
  A preference optimization framework for code models that focuses on both correctness and efficiency without depending on external resources.

• FocusedDPO | ACL 2025 Finding
  Fine-graind optimization for code models with Error-Prone Points Identification.

• CodeAgent | ACL 2024 Main Conference
  Integrating multiple programming assistance tools into large models for practical problem-solving.

• HiRoPE | ACL 2024 Main Conference
  Introducing a plug-and-play length extension method for large code models.

• Self-Edit | ACL 2023 Main Conference
  Early exploration into the self-repair capability of large models in code generation.

• Hierarchy Transformer | ICPC 2023 & EMSE 2024 | 🏆ACM SIGSOFT Distinguished Paper Award
  A novel Transformer structure for modeling both sequence and structural information in source code.

  We further pretrain the 125M code model with this new architecture.

• Heterogeneous Code GNN | ICPC 2022
  Proposing a heterogeneous graph representation model for programs.

• ToolCoder
  Tool-enhanced learning method embedding external API search tools into code generation models.

• Code Generation Survey | SCIS 2024, CCF-A
  A comprehensive survey of code generation.

---

🏆 Honors & Awards

• 2023 ACM SIGSOFT Distinguished Paper Award
• Peking University Outstanding Student Award (2022, 2023)
• 2023 Peking University Yongying Foundation Scholarship
• Peking University Excellent Research Award (2017-2021)
• Peking University EECS Scholarship (2017-2021)
• 2020 Peking University Schlumberger Scholarship

---

📬 Contact Information

• Email: zhangkechi@pku.edu.cn
• Homepage: Google Scholar
• Address: Room No. 1726, No. 1 Science Building, Peking University, No. 5 Yiheyuan Road, Haidian District, 100871 Beijing

---

Feel free to explore my publications.

---