2025

TadReamk Limited, Hong Kong
Jun. 2025 – Aug. 2025 • Internship on Research and Development (Research focused) | Intern
  • Logo Layout & SVG Generation: Trained transformer models for logo layout generation and text-to-SVG conversion; developed B-spline-based algorithm to convert raster logos into vector primitives.
  • Preference Selection System: Deployed a pairwise preference selection system using info-gain active sampling with approximate message passing algorithm for ranking logo quality.
1TadReamk Limited, Hong Kong

Jun. 2025 – Aug. 2025

3

  • Logo Layout & SVG Generation: Trained transformer models for logo layout generation and text-to-SVG conversion; developed B-spline-based algorithm to convert raster logos into vector primitives.
  • Preference Selection System: Deployed a pairwise preference selection system using info-gain active sampling with approximate message passing algorithm for ranking logo quality.

Tackling Multi-user Alignment Defects in XR Software Using Multi-Agent System
Prof. Michael R. Lyu • Jun. 2025 - present • Bachelor's Thesis
  • Multi-Agent System Development: Developed a multi-agent system with the PocketFlow framework, integrating RAG and message queues for XR software testing automation.
  • Inter-Agent Collaboration: Designed and implemented inter-agent collaboration workflows for efficient task coordination.
  • Unity Package Development: Created a configurable Unity package for automated multi-view data capture and camera parameter extraction in avatar-centered scenes.
1Tackling Multi-user Alignment Defects in XR Software Using Multi-Agent System

Jun. 2025 - present

3

  • Multi-Agent System Development: Developed a multi-agent system with the PocketFlow framework, integrating RAG and message queues for XR software testing automation.
  • Inter-Agent Collaboration: Designed and implemented inter-agent collaboration workflows for efficient task coordination.
  • Unity Package Development: Created a configurable Unity package for automated multi-view data capture and camera parameter extraction in avatar-centered scenes.

2024

Designing AI Agent System in VR Software Testing
Shuqing Li • Sep. 2024 – Dec. 2024 • Research Assistant
  • Designing an AI Agent for automated VR game testing and assisting in developing the task execution framework.
  • Learning and implementing Retrieval-Augmented Generation (RAG) and Model-Based Testing frameworks for the Agent.
1Designing AI Agent System in VR Software Testing

Sep. 2024 – Dec. 2024

3

  • Designing an AI Agent for automated VR game testing and assisting in developing the task execution framework.
  • Learning and implementing Retrieval-Augmented Generation (RAG) and Model-Based Testing frameworks for the Agent.

ComboBench: Can LLMs Manipulate Physical Devices to Play VR Games? [CVPR Under Review]
Prof. Michael R. Lyu • Apr. 2024 – Aug. 2024 • Summer Research Internship
  • Benchmark Implementation: Curated and structured 270 VR game tasks from four top-rated VR games, establishing the first benchmark for evaluating LLMs performance in immersive VR environments.
  • Experiment: Benchmarked 7 cutting-edge LLMs on VR tasks, assessing their effectiveness in completing complex game objectives.
  • Data Analysis: Conducted in-depth data analysis, assisting in the development of 3 scoring systems for robust LLMs performance evaluation in VR settings.
  • Survey: Designed and distributed questionnaires using Qualtrics to gather human evaluations of VR tasks, comparing them to AI model results.
1ComboBench: Can LLMs Manipulate Physical Devices to Play VR Games? [CVPR Under Review]

Apr. 2024 – Aug. 2024

3

  • Benchmark Implementation: Curated and structured 270 VR game tasks from four top-rated VR games, establishing the first benchmark for evaluating LLMs performance in immersive VR environments.
  • Experiment: Benchmarked 7 cutting-edge LLMs on VR tasks, assessing their effectiveness in completing complex game objectives.
  • Data Analysis: Conducted in-depth data analysis, assisting in the development of 3 scoring systems for robust LLMs performance evaluation in VR settings.
  • Survey: Designed and distributed questionnaires using Qualtrics to gather human evaluations of VR tasks, comparing them to AI model results.

2023

An AI-enhanced Adaptive and Individualized eLearning System for Mathematics Foundation Courses in the Faculty of Engineering
Dr. Dongkun Han • Jun. 2023 – Sep. 2023 • student helper
  • Data Collection: Collected, organized, and managed comprehensive background data on Hong Kong secondary schools and students.
  • Model Implementation: Assisted in implementing different classification algorithms for predicting students’ learning levels.
  • Model Tuning: Assisted in tuning parameters and organizing training data to enhance model performance.
1An AI-enhanced Adaptive and Individualized eLearning System for Mathematics Foundation Courses in the Faculty of Engineering

Jun. 2023 – Sep. 2023

3

  • Data Collection: Collected, organized, and managed comprehensive background data on Hong Kong secondary schools and students.
  • Model Implementation: Assisted in implementing different classification algorithms for predicting students’ learning levels.
  • Model Tuning: Assisted in tuning parameters and organizing training data to enhance model performance.