Dr. Jacky Wai Keung
Director 6204 Scopus citationsh-index 41Stanford top-2% 2022 & 2023IEEE Hong Kong CS Chairman13+ research awardsHK$20M+ funding290 publications Accepting PhD applicants and open to industry collaboration. Personal website ↗Google Scholar ↗CityU Scholars ↗About
Prof. Jacky Wai Keung leads the Agentic Software Intelligence (ASI) Lab at City University of Hong Kong, where he is a Professor in the Department of Computer Science. His work sits at the intersection of large language models, agentic and multi-agent systems, and empirical software engineering.
He received his B.Sc. (Hons) in Computer Science from the University of Sydney and his Ph.D. in Software Engineering from the University of New South Wales, and has held research roles in the Software Systems Research Group at NICTA (now DATA61/CSIRO). His work has produced over 290 peer-reviewed publications and more than HK$20M in funded research across GRF, ITF and TDG initiatives, and he has been named a Stanford top-2% most-cited scientist (2022 and 2023).
Prof. Keung is dedicated to making learned, self-directed software systems reliable, secure and trustworthy across real-world ecosystems. He actively recruits PhD students and collaborates with industry.
Research Interests
Publications (251)
- Where Do Expectations Diverge? an Empirical Analysis of Software Engineering Graduate Preparedness in Industry.
- When Fine-Tuning LLMs Meets Data Privacy: An Empirical Study of Federated Learning in LLM-Based Program Repair.
- UniAda: Universal Adaptive Multi-objective Adversarial Attack for End-to-End Autonomous Driving Systems.
- Understanding How Enterprises Adopt the Model Context Protocol for LLM-Driven Software Engineering.
- To Answer or to Abstain: Mitigating Search-Agent Hallucinations via Abstention-Aware Reinforcement Learning.
- Teaching Software Engineering with LLM and MCP Integration: From Classroom to Industry Practice.
- SysTradeBench: An Iterative Build-Test-Patch Benchmark for Strategy-to-Code Trading Systems with Drift-Aware Diagnostics.
- Refine After Generation: Toward Correct and Concise Patches in LLM-based Program Repair.
- R2ComSync: improving code-comment synchronization with in-context learning and reranking.
- R2Code: A Self-Reflective LLM Framework for Requirements-to-Code Traceability.
- NFR-to-Code Traceability in a Blockchain-IoT System: An Empirical Study.
- Making Sense of Scams: Understanding Scam Conversations Through Multi-Level Alignment.
- LogMeta: A few-shot model-agnostic meta-learning framework for robust and adaptive log anomaly detection.
- LaViT: Aligning Latent Visual Thoughts for Multi-modal Reasoning.
- Industrial log analysis revisited: A task-oriented evaluation of parsing and anomaly detection under real-world constraints.
- Improving anomaly detection in software logs through hybrid language modeling and reduced reliance on parser.
- HumanEval-V: Systematic Evaluation of Visual Reasoning in Large Multimodal Models for Code Generation
- HGARena: Budgeted Test-Driven Multi-Agent Repository Issue Resolution with Heterogeneous Graph-Augmented Retrieval.
- FedDC: Efficient protection scheme based on chaotic system in federated learning.
- Empirical Insights of Test Selection Metrics under Multiple Testing Objectives and Distribution Shifts.