CV

Education

Hong Kong University of Science and Technology (Guangzhou), AI Thrust
MPhil in Artificial Intelligence, 2025-09 - 2027-07 (expected)
GPA: 4.07/4.3; full scholarship

  • Working with Prof. Chengwei Qin on AI long-term memory, multi-agent orchestration, and personalized agents.
  • Relevant coursework: Advanced Machine Learning, Reinforcement Learning, Natural Language Processing, and Distributed Systems.

Hong Kong Baptist University
BSc(Hons) in Data Science, 2021-09 - 2025-06
GPA: 3.59; First Prize Scholarship

Publications

Conference papers

  • T-Mem: Memory That Anticipates, Not Archives. EMNLP 2026 Main Conference (accepted). Paper · Code
  • Unified-MAS: Universally Generating Domain-Specific Nodes for Empowering Automatic Multi-Agent Systems. EMNLP 2026 Main Conference (accepted). Paper · Code
  • AMA: Adaptive Memory via Multi-Agent Collaboration. Co-first author; Findings of ACL 2026. Paper · Code
  • BizCompass: Benchmarking the Reasoning Capabilities of LLMs in Business Knowledge and Applications. Findings of ACL 2026. Paper
  • UNGIA: Unnoticeable Injection Attack on Graph Neural Networks. IJCNN 2026.

Public preprints

  • Beyond Corrected Memory: Execution Consistency in Multi-Agent Systems. Co-first author; arXiv:2610.08101; submitted to ICLR 2027. Paper
  • MAS-Gate: Unlocking Multi-Agent System Lifecycle via Node-Centric Quantification. Co-first author; arXiv:2610.17613; submitted to ICLR 2027. Paper
  • Degrees of Freedom in Multi-Agent Coordination: Between Centralized and Decentralized Regimes. arXiv:2610.17751; submitted to ICLR 2027. Paper

Research and Work Experience

Applied Research Algorithm Engineer Intern, Tencent, 2026-04 - 2026-09

  • Co-developed T-Mem, an associative long-term memory approach accepted to EMNLP 2026 Main Conference.
  • Built a retrieval and LLM reranking pipeline that improved exact-hit rate from 60.3% (38/63) to 84.1% (53/63) on an offline set using only pre-query history.
  • Built annotation tooling and quality-checked topic-matching and false-recall evaluation sets to guide retrieval and response improvements.

Research Assistant, Guangdong Province Key Laboratory of Interdisciplinary Research and Application for Data Science, Zhuhai, China, 2024-06 - 2024-09

  • Developed a model architecture supporting incremental learning based on the Learning to Edit long-term memory mechanism.
  • Assisted in building LLM agents for medical scenarios, integrating text, image, and structured medical-record data.
  • Used LangChain to design workflows for medical agents and combined Spark for distributed preprocessing and privacy-protection work.

Data Analyst Manager, Vivo Mobile Communication Co., Ltd., Shenzhen, China, 2023-06 - 2023-09

  • Completed 20+ analyses, generated 40+ insights and recommendations, and supported 10+ initiatives that cumulatively drove an increase of 110,000 daily active users.
  • Built dashboards accessed more than 800 times by colleagues.
  • Participated in the framework training of the company's data-assistant LLM model and developed a Python-based demo.
  • Helped improve single-table SQL generation accuracy from around 45% to 62-67%.

Selected Research Projects

Research on the Stealthiness of Injection Attacks in Graph Neural Networks, undergraduate thesis supervised by Prof. Donglong Chen, 2024-06 - 2024-09

  • Proposed UNGIA, a stealthy graph injection attack method using clustering techniques to identify representative nodes.
  • Conducted experiments on Cora, Citeseer, and Cora-ML.

Chaos-Based GNN for Aspect-Based Sentiment Analysis, collaborative project, 2023-11 - 2024-09

  • Explored models for polysemy, semantic correlation, and emotion complexity in social-media sentiment analysis.

CNO-former: Chaotic Neural Oscillation Self-Attention Model, research project, 2023-09 - 2024-09

  • Formulated a Transformer-style model based on chaotic neural oscillation and self-attention.
  • Awarded an Excellent Award in the 2nd Computer Science Related Poster Exhibition.

Advanced Multi-Agent Systems Development, Guangdong Provincial Special Scientific Research Program, 2023-05 - 2025-02

  • Worked on a multi-agent teaching community system based on reinforcement learning.
  • Developed a multi-agent tour-guide assistance system for personalized tourism services.

Competitions and Honors

  • Bronze Medal, Kaggle LLM Prompt Recovery Competition, 2024.

  • Meritorious Winner, Mathematical Contest in Modeling, 2023.
  • Honorable Mention, Mathematical Contest in Modeling, 2022.

  • First Prize Scholarship, Hong Kong Baptist University.