SELAB@HKNU

2024: Applied Small Large Language Models (sLLMs) and Large Multimodal Models (LMMs) to software engineering, focusing on the automated generation and optimization of software artifacts. This approach enhanced the adaptability and scalability of legacy systems through automated refactoring and code optimization techniques.
2024: Integrated Retrieval-Augmented Generation Small Large Language Models (RAG-sLLMs) to dynamically incorporate external data sources, improving precision in generating context-aware software solutions. RAG-sLLMs provided adaptable responses for complex software engineering tasks, making them essential for advanced applications.
2024: Utilized Large Language Models (LLMs), such as GPT and BERT, to streamline software artifact generation and code optimization. These LLMs played a key role in automating tasks like code completion, documentation, and natural language processing within software development.
2024: Implemented Small Language Models (SLMs) with a focus on efficiency and scalability, deploying these resource-efficient models for specific software development tasks that required fast processing and reduced computational demands.
2024: Incorporated Large Multimodal Models (LMMs) within software engineering workflows to expand model capabilities beyond text. LMMs leveraged their understanding of multiple modalities, including natural language, images, and structured data, to improve analysis, debugging, and overall software development processes.

2023: Enhanced Automatic Program Repair using CodeBERT and Call Graph-Based Learning, showing significant improvements in fixing software bugs autonomously.
2022: Advanced Bug Report Duplication Prediction through topic-Based Learning and a Fine-Tuned BERT Algorithm, reducing redundancy in bug tracking systems.
2022: Innovated Bug Triage with Developer-Focused Feature Selection and a CNN-LSTM Approach, improving the efficiency of bug assignment processes.
2022: Predicted Bug Severity using topic-Based Feature Selection combined with the CNN-LSTM Algorithm, aiding in prioritizing critical software issues.
2021: Improved Bug Localization by analyzing similar commits with topic-Based insights using CNN-LSTM, enhancing the accuracy of identifying buggy code segments.
2020: Leveraged SeqGAN for effective software bug repair, showing the potential of generative adversarial networks in software maintenance.

2023: Developed Neural Network Techniques for analyzing unstructured medical records and disease classification, providing advancements in healthcare data processing.

2023: Created personalized voice synthesis assistive algorithms for the disabled, powered by neural networks, offering significant improvements in assistive technologies.

2023: Predicted emotions in free speech with a multi-modal approach to voice and text relevance learning, contributing to more nuanced human-computer interactions.
2022: Identified voice phishing with a deep learning algorithm integrating smishing, voice phishing, and anomaly detection, enhancing security measures against fraudulent activities.

2023: Developed hybrid deepfake detection and creation technologies using global and local feature analysis, advancing the field of digital media authentication and synthesis.

2023: Estimated sleep quality with algorithms analyzing the relationship between lifelogs and sleep data, contributing to health monitoring and improvement technologies.
2022: Developed an intelligent StarCraft 2 agent using reinforcement learning, pushing the boundaries of AI in gaming.
2022: Created a fake news detection algorithm utilizing neural networks, addressing the critical issue of misinformation in digital media.

AI-Powered Software Engineering (SELab)