Sergey Mechtaev / 谢辽夏

Sergey Mechtaev's Portrait
Email mechtaev@pku.edu.cn
GitHub github.com/mechtaev
Twitter x.com/sergey_mechtaev
Address Room 1542, Science Building #1, Peking University, No. 5 Yiheyuan Road, Beijing, 100871, China

Sergey Mechtaev is an Assistant Professor at Peking University. Previously, he obtained a PhD degree from the National University of Singapore, and was a faculty member at University College London. His research aims to automate the development and maintenance of trustworthy software. It has been recognized by prestigious awards, notably, an ACM SIGSOFT Outstanding Dissertation Award and two ACM SIGSOFT Distinguished Paper Awards, and has been deployed in industry, at Bloomberg, and academia, at IIT Kanpur.

Publications

  • Reducing Hallucinations in LLM-Generated Code via Semantic Triangulation. Yihan Dai, Sijie Liang, Haotian Xu, Peichu Xie, Sergey Mechtaev. Draft 2026.

  • Statistical Independence Aware Caching for LLM Workflows. Yihan Dai, Dimitrios Stamatios Bouras, Haoxiang Jia, Sergey Mechtaev. LLM4Code@ICSE 2026.

  • HoarePrompt: Structural Reasoning About Program Correctness in Natural Language. Dimitrios Stamatios Bouras, Yihan Dai, Tairan Wang, Yingfei Xiong, Sergey Mechtaev. ICSE 2026.

  • Automated Repair of Ambiguous Problem Descriptions for LLM-Based Code Generation. Haoxiang Jia, Robbie Morris, He Ye, Federica Sarro, Sergey Mechtaev. ASE 2025.

News

30/11/25 A paper on caching LLM workflows has been accepted to LLM4Code@ICSE'26.
15/10/25 A paper on LLM reasoning about program correctness has been accepted to ICSE'26.
25/09/25 A paper on repairing ambiguous requirements has been accepted to ASE'25.
04/02/25 A paper on LLM for genetic programming has been accepted to GI@ICSE'25.
23/01/25 I gave a talk on synergizing program analysis and ML at Shonan Meeting 217.
31/10/24 A paper on selecting relevant information for LLM-based program repair has been accepted to ICSE'25.
01/08/24 I started working as an Assistant Professor at Peking University.
02/07/24 A paper on LLM-based code generation has been accepted to ISSTA'24.
01/07/24 I resigned from UCL and received an honorary lecturer title.

Group

  • Yihan Dai
    PhD Student at PKU

  • Dimitrios Stamatios Bouras
    PhD Student at PKU

  • Haoxiang Jia
    PhD Student at PKU

Talks

December 2025 Reducing Hallucinations in LLM-Generated Code via Semantic Triangulation. Zhejiang University.
June 2025 Quality Assurance in The Era of Natural Language Driven Programming. Shanghai Workshop on Structural AI.
May 2025 Quality Assurance in The Era of Natural Language Driven Programming. Shanghai Jiao Tong University.
May 2025 Quality Assurance in The Era of Natural Language Driven Programming. Fudan University.

Service

  • ICSE 2026
    Program committee

  • LMPL@SPLASH 2025
    Organizer

  • APR@ICSE 2025
    Organizer

  • ICSE 2025
    Program committee

  • FSE 2025
    Program committee