What's New
About Me
I am a first-year Ph.D. student at Beihang University, working in the SMAT Laboratory under the supervision of Prof. Li Li (黎立), co-supervised by Mingyi Zhou (周鸣一).
My research focuses on AI for Software Engineering (AI4SE), combining Large Language Models (LLMs), static analysis, and verifiable software-engineering environments to study reliable coding agents and ecosystem-specific toolchains.
Research Directions
I collaborate closely with Huawei on OpenHarmony quality and UI performance, Wei Ma at BTH on LLM code understanding and AI4SE evaluation, and Zhensu Sun at SMU on LLM coding agents and execution efficiency. My work has been published at top SE venues including ICSE, FSE, TOSEM, and EMSE.
Research Interests
🔬 Research Directions
Coding Agents for Software Change
Measuring how coding agents choose context, execution, repository relations, and history organization when making real code changes.
- Execution, static anchors, impact relations, and commit structure are measurable design choices
- Process quality matters alongside final patch correctness
Repository-Scale Benchmarks and Verifiable SWE Environments
Building low-leakage, execution-backed tasks where agents must repair, migrate, or rescue real repositories under observable constraints.
- Whole-repository tasks expose failures that file-level benchmarks miss
- Source-only checks, runtime blocking, and scenario validation make evaluation harder to game
OpenHarmony and Emerging-Language Toolchains
Building ecosystem-aware analysis, repair, migration, performance, and program-analysis infrastructure for OpenHarmony, ArkTS, and Cangjie.
- From app repair and UI performance to Android migration and OpenHarmony-specific static analysis
- Emerging ecosystems make domain knowledge, executable feedback, and low-resource adaptation central
Code Understanding, Program Analysis, and AI4SE Measurement
Evaluating what LLMs understand about code and designing measurement protocols for AI-based SE tools and LLM-in-the-loop systems.
- Capability claims should be tied to syntax, static behavior, dynamic behavior, and tool failure modes
- Evaluation protocols need human audit, leakage control, and failure-mode analysis
Open-Source Tools
HapRay
Performance analysis tool for OpenHarmony apps, detecting Phantom Rendering issues in UI.
HomeCheck
Static checker and auto-repair tool for OpenHarmony apps. Core module of HapRepair with rule-driven defect detection.
HomeTrans
Android → OpenHarmony migration tool, helping developers automatically convert Android apps to the HarmonyOS ecosystem.
ResearchClaw
⭐ 59AI-powered research desktop app combining literature management, smart reading notes, and research idea generation. Supports arXiv discovery, PDF AI chat, semantic search, and citation networks.
SkillLens
Local SKILL coverage and optimization framework for Codex / Claude Code traces, aligning SKILL.md constraints with real agent behavior to reveal followed, violated, and ignored paths.
Accepted Papers (PDF)
View all →To Run or Not to Run: Analyzing the Cost-Effectiveness of Code Execution in LLM-Based Program Repair
ISSTA 2026
How Much Static Structure Do Code Agents Need? A Study of Deterministic Anchoring
ISSTA 2026
Phantom Rendering Detection: Identifying and Analyzing Unnecessary UI Computations
FSE 2026
MazeBreaker: Multi-Agent Reinforcement Learning for Dynamic Jailbreaking of LLM Security Defenses
ICSE 2026
HapRepair: Learn to Repair OpenHarmony Apps
FSE Industry 2025
Effective Fine-tuning for Low-resource Languages: A Case Study of Cangjie
EMSE 2026
Open-Source AI-based SE Tools: Opportunities and Challenges of Collaborative Software Learning
TOSEM 2024
Exploring Code Analysis: Zero-Shot Insights on Syntax and Semantics with LLMs
TOSEM 2026
Collaborations
- HapRepair: OpenHarmony app repair
- Cangjie low-resource language fine-tuning
- Phantom Rendering detection for mobile UI performance
- Phantom Rendering detection for mobile UI performance
- Phantom Rendering detection for mobile UI performance
- • OpenHarmony app repair
- • Phantom Rendering / UI performance analysis
- • Cangjie low-resource language fine-tuning
Heartfelt thanks to Wei Ma for guiding me into research. When I first began, he patiently helped me with topic selection, paper reading, experiment design, and writing; much of my early research training was shaped by his guidance and support.
- MazeBreaker: multi-agent RL for LLM jailbreak evaluation
- HapRepair: LLM-guided repair for OpenHarmony apps
- Exploring Code Analysis: syntax and semantic probing with LLMs
- Open-source AI-based SE tools survey
- • LLM code understanding and semantic evaluation
- • MazeBreaker adaptive evaluation
- • AI4SE open-source ecosystem studies
Heartfelt thanks also to Zhensu Sun. After I became able to develop research ideas more independently, he continued to offer constructive advice on problem framing, experiment design, paper narrative, and limitation analysis, helping me learn how to refine early ideas into more complete and solid research work.
- EAGER: executing code as LLMs generate it
- To Run or Not to Run: execution cost-effectiveness in LLM repair agents
- • LLM coding agents
- • Execution cost and latency optimization
- • Code generation and program repair evaluation