Yongxin Guan (Kevin)

Yongxin Guan (Kevin)

B.Sc. (Hons) in Data Science @ BNBU

I am Yongxin Guan (Kevin), a Year 3 student in the Bachelor of Science (Honours) in Data Science program at Beijing Normal-Hong Kong Baptist University (BNBU). My long-term goal is to pursue a PhD and build research that connects multimodal intelligence with practical deployment in real-world systems.

My current interests lie in AI for Science, world models, knowledge graphs, agent systems, reinforcement learning, and LLM enhancement across different domains. I enjoy working on problems that combine language, vision, structured knowledge, and robust engineering.

I am currently conducting research on multimodal relation extraction with large language models, knowledge graph completion for Graph RAG, and vision-based measurement systems. I am also open to research collaboration, internships, and opportunities where I can contribute to AI product building and applied machine learning.

My CV is available for download. You can also find me on GitHub and Kaggle.

🔥 News

  • 2026.08:  🎉 Our paper “TrustIE: Trustworthy Information Extraction via Joint Uncertainty Estimation and Evidence Attribution” is accepted by ICONIP 2026 (CCF-C).
  • 2026.08:  Our team won the National Third Prize in the 12th National College Students Statistical Modeling Competition with the paper “中国省域算电协同的水平测度、时空演化与政策效应评估——基于耦合协调度、空间杜宾模型与PSM-DID的实证研究”.
  • 2026.03:  🎉 Our paper “Enhancing Multimodal Adversarial Transferability via Boundary-Context Augmentation” is accepted by ICCECT 2026.
  • 2026.06:  Started as a Research Assistant at the Big Data Institute, HKUST-GZ, advised by Prof. Lei Chen.
  • 2026.03:  Started as a Visiting Internship Student at VisLab, HKUST, advised by Prof. Huamin Qu.
  • 2025.08:  Selected as a Fellow of TRAE, ByteDance.

📝 Research & Publications

ICCECT 2026 (Accepted)
sym

Enhancing Multimodal Adversarial Transferability via Boundary-Context Augmentation

ICCECT 2026 (Accepted) Co-author AI Security
  • Proposed the BCA-AET framework to improve adversarial robustness under different spatial layouts through boundary-context augmentation.
  • Integrated image-text collaborative optimization to strengthen cross-modal attacks and improve transferability for VLP models in black-box settings.
ICONIP 2026 (Accepted)
sym

TrustIE: Trustworthy Information Extraction via Joint Uncertainty Estimation and Evidence Attribution

ICONIP 2026 (Accepted) First Author CCF-C
  • 2026.08: Accepted.
  • FYP: Multimodal Relation Extraction with Large Language Models (Core Member, Supervisor: Dr. Rui Meng, Mar 2026 - Present). Designed multimodal prompt strategies for LLMs to fuse visual and textual features for relation extraction; built and pre-processed multimodal datasets and aligned image-text feature spaces. (Ongoing)
  • Knowledge Graph Completion and Graph RAG (Recent Research Topic, Supervisor: Dr. Rui Meng). Evaluated and implemented KGC methods for link prediction and entity alignment; used KGC to strengthen Graph RAG pipelines, reduce hallucination, and improve reasoning quality in language models.
  • Height Measurement Algorithm for Human Talent Identification and Development (Research Project, Supervisor: Dr. Ximing Shang). Developed a single-image visual algorithm for high-precision height measurement with human keypoint extraction, distortion and perspective correction.

🎖 Honors and Awards

  • 2026 12th National College Students Statistical Modeling Competition - National Third Prize
  • 2026 The 6th Meituan Business Analysis Elite Competition - Outstanding Works Award (Global Top 50-100)
  • 2025 The 16th National College Students’ E-commerce Innovation, Creativity and Entrepreneurship Challenge Competition - First Prize
  • 2025 2025 National College Students Mathematical Modeling Competition - Provincial Third Prize
  • 2025 2025-2026 Student Leadership Development Program - Silver Medal
  • 2025 TRAE Shenzhen Hackathon - TRAE SOLO Best Idea Award
  • 2025 2024-2025 School-level Second-Class Scholarship
  • 2024 Environmental Awareness Holistic Education Outstanding Student Award

📖 Educations

🏛 Research Appointments

  • 2026.06 - Present, Research Assistant, Big Data Institute (BDI), HKUST Guangzhou, advised by Prof. Lei Chen. Developing robust, deployment-oriented agent system applications for real-world scenarios, with a focus on reliability, stability, and consistent task execution under noisy environments.
  • 2026.03 - Present, Visiting Internship Student, VisLab, HKUST, advised by Prof. Huamin Qu. Exploring the integration of agentic workflows with domain-specific languages and structured procedural representations.

💻 Internships

  • 2026.01 - 2026.03, Agent Engineer Intern, China Unicom. Built an intelligent agent for PDF content analysis of image-heavy enterprise documents, integrating OCR-based visual and textual parsing.
  • 2025.08 - Present, Fellow, TRAE, ByteDance. Evaluate AI-driven development tools, conduct product testing, and analyze user needs to improve developer experience.
  • 2025.06 - 2025.08, Backend Development Engineer Intern, China Mobile. Built enterprise business platforms with Dify-based workflows, backend services, database logic, and APIs.
  • 2025.08 - Present, Campus Intern, Career Development Center, BNBU. Manage and analyze student employment data and coordinate campus recruitment operations.

🌟 Leadership & Campus Service

  • Navigator Robot Team of BNBU: Member of the Mechanical Development Group in the VEX U and RoboMaster team; optimized robot structures through mechanical analysis and represented the college in competitions.
  • Chinese Communist Youth League Affairs Officer: Propaganda Department Officer Member at BNBU; produced event posts and communication materials.
  • Badminton Association of BNBU: Propaganda Department Officer and Minister; coordinated resources and supported event organization.

🛠 Skills & Interests

  • Programming: Python, PyTorch, OpenCV, Java, C, SQL, TypeScript
  • Technical Strengths: Full-stack development, data analytics, AI tools, paper writing, multimodal system prototyping
  • Languages: Cantonese, English, Putonghua, Spanish (Beginner), French (Beginner)
  • Interests: AI research, badminton, photography, books, aviation and bus culture, making and tasting coffee