Personal Information

Name: Kechi Zhang (张克驰)
Address: Room No. 1726, No. 1 Science Building, Peking University, No. 5 Yiheyuan Road, Haidian District, 100871 Beijing
Email: zhangkechi@pku.edu.cn
Homepage: Google Scholar

Education

  • Ph.D.
    School of Computer Science, Peking University, Beijing, China
    Duration: Sept. 2021 - July 2026 (expected)
    Major: Computer Software and Theory – Intelligent Software and Requirement Engineering
    Tutor: Professor Zhi Jin, Professor Ge Li

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

Research Direction

AI4SE & LLMs for Code

  • Code Generation and Code Representation Based on Deep Learning
    • Pre-training, fine-tuning, and alignment of Code LLMs
    • Tool Enhancement, Agent Technology, and Length Extrapolation for Code Models
    • Project-level Code Generation
    • Code Representation Model based on Structural Information

Main Research Papers

  • CodeAgent
    ACL 2024 Main Conference
    • Code generation work proposing a method to integrate multiple programming assistance tools into large models to solve practical programming problems
  • HiRoPE
    ACL 2024 Main Conference
    • Work on length extrapolation in large code models, proposing a plug-and-play length extension method that requires no training
  • Self-Edit
    ACL 2023 Main Conference
    • Code generation work, one of the earliest explorations into the self-repair capability of large models in code generation
  • Hierarchy Transformer
    ICPC 2023
    🏆 ACM SIGSOFT Distinguished Paper Award 🏆
    • Code representation work, proposing an improved Transformer structure for jointly modeling sequence information and structural information in source code
  • Heterogeneous Code GNN
    ICPC 2022
    • Code representation work, proposing a heterogeneous graph representation model for programs, HGT-HPG, to model the graph structure information in code
  • ToolCoder
    • Code generation work proposing a tool-enhanced learning method that embeds external API search tools (internet search engines and document search tools) into the code generation model
  • Code Generation Survey
    SCIS 2024, CCF-A
    • A comprehensive survey of code generation, summarizing related work in the field

Rewards & Honors

  • 2023 ACM SIGSOFT Distinguished Paper Award
  • Peking University Outstanding Student Award, 2022 and 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