CV
Academic CV of Lishui Fan.
Contact Information
| Name | Lishui Fan |
| flscode@zju.edu.cn |
Experience
-
2026 - present China
Research Intern
Baidu ERNIE Bot Team
Working on large language model applications and code intelligence for ERNIE Bot.
-
2026 - 2026 China
Research Intern
ByteDance Douyin AI4SE
Worked on AI for software engineering and code intelligence problems in practical development settings.
-
2025 - 2025 China
Research Intern
Taotian Future Life Lab
Conducted applied AI research for future-life and intelligent systems.
Education
-
2023 - present Hangzhou, China
Ph.D. Student
Zhejiang University
Computer Science and Technology
- College of Computer Science and Technology.
- Research focus: LLM-based code generation and software engineering.
-
2020 - 2023 Nanjing, China
M.Eng.
Nanjing University
Software Engineering
- State Key Laboratory for Novel Software Technology.
-
2016 - 2020 Tianjin, China
B.Eng.
Tianjin University
Software Engineering
- GPA: top 10%.
Research
- LLM-based Code Generation — training, evaluation, and optimization methods for code LLMs, including fault-guided fine-tuning, keyword-enhanced prompting, and reasoning-process reward learning.
- Code-Change Understanding — systematic studies of LLM capabilities on code review generation, commit message generation, and just-in-time comment update.
- Industry Collaboration — core contributor to projects with Huawei and Tencent (CCF-Huawei Populus Grove Fund, Tencent Rhino-Bird special project), translating practical software engineering needs into model training, evaluation, and optimization problems.
Publications
-
2026 ReCode: Reinforcing Code Generation with Reasoning-Process Rewards
ACL 2026 Main Conference
A CCF-A paper on reasoning-process reward learning for code generation.
-
2026 ZeroCoder: Can LLMs Improve Code Generation Without Ground-Truth Supervision?
arXiv
A preprint on label-free co-evolution of code and test generation.
-
2026 Balancing Latency and Accuracy of Code Completion via Local-Cloud Model Cascading
FSE 2026
A collaborative Tencent-related paper on local-cloud model cascading for code completion.
-
2025 Exploring the Capabilities of LLMs for Code-Change-Related Tasks
ACM Transactions on Software Engineering and Methodology
A CCF-A journal article studying LLMs for code-change-related tasks.
-
2025 FGit: Fault-Guided Fine-Tuning for Code Generation
ASE 2025
A CCF-A conference paper on fault-guided fine-tuning for code generation.
-
2025 SEK: Self-Explained Keywords Empower Large Language Models for Code Generation
Findings of ACL 2025
A CCF-A paper on keyword-enhanced code generation with LLMs.
Projects
-
CCF-Huawei Populus Grove Fund Project
Core research contributor to an industry-academic project on software engineering intelligence.
- Code LLM training and evaluation.
- Translation of real software engineering needs into research problems.
-
Tencent Rhino-Bird Special Project
Core contributor to a Tencent collaboration on code intelligence and code completion.
- Related collaborative work accepted to FSE 2026.
- Focus on latency-accuracy trade-offs for code completion.
-
Huawei Industry Collaboration Project
Core research contributor to a Huawei industry collaboration project.
- Practical software engineering scenarios.
- Code LLM optimization and evaluation.