Bowen Xu
Assistant ProfessorSoftMax Lab
Software Engineering Group at NCSU
Department of Computer Science
College of Engineering
Office: Engineering Building II (EB2), Room 3228
Email: bxu22@ncsu.edu
My team has developed a framework named Evaloop to fairly assess LLMs' robustness for programming. Based on this, we are actively mainteining a leaderboard with the results on more than 100 LLMs. Check it out at https://evalooop.github.io!
Research Focus
My research interests span software engineering, machine learning, and program analysis. My primary research interests are:- AI for Code
- Privacy Risk [AIWare-25]
- Backdoor Attack and Defense on Large Code Models [TSE-24]
- Code Data Interpretation and Quality [ASE-23] [ICPC-22]
- Code Representation Learning [TOSEM-23] [ICSME-23] [SANER-22] [FSE-22] [TSE-21]
- Model Compression [ASE-22] [ICSE-SEIS-24] [ICSME-25]
- Software Vulnerability
- Vulnerability Detection [ASE-23]
- Vulnerability Repair [ICSE-24] [TSE-23] [AIWare-24]
- Safety of AI-enabled Software System
- Violation Identification of AI System [TOSEM-25]
- Backdoor Attack in Reinforcement Learning [S&P-24]
- Understand and Assist Developers' Activities
- Code Reuse and Re-implementation [EMSE-19]
- Automatic Code Review [ICPC-23] [PAKDD-19]
- Bug Management and Analysis [TOSEM-23] [TSE-25]
- Software Knowledge Interpretation [EMSE-24] [ASE-23] [ESEM-18] [ASE-16]
- Developer-oriented Chatbot [ICSE-23] [ASE-22] [FSE-19] [ASE-17][FSE-17] [EMSE-17] [MSR-16]
- Semantic Code Search [FSE-22]
September, 2025: One work titled How Quantization Impacts Privacy Risk on LLMs for Code? is accepted by Main Track of 2nd ACM International Conference on AI-powered Software (AIware 2025). Congrats Nazmul and Hua!
