Xiaofeng Li
Undergraduate in Artificial Intelligence · LLM evaluation & agents
I am an undergraduate student in Artificial Intelligence at the Hong Kong University of Science and Technology (Guangzhou), graduating in 2027 (BEng, CGPA 3.6+/4.3, top 15%). My work centers on evaluating and understanding large models — building evaluation systems that align with human perception, and applying agents to data synthesis and automation.
Research Interests
- Large Language Models — evaluation and trustworthiness: benchmark design, evaluation methodology, and evidence-grounded verification
- Agents — long-horizon task solving, tool use, and evaluation-driven data synthesis and automation
Research & Projects
- Image-Caption Evaluation Agent (Developer, Prof. Jiaheng Wei’s group · Huawei collaboration, ongoing) — I design the caption-precision module of a high-quality image-text data synthesis pipeline. It is the first use of an agent system for this task: the agent understands an image hierarchically from coarse to fine, aligned with human perception, and reconstructs a scene graph, enabling reference-free caption evaluation. See the project page.
- Long-Horizon Agent Trajectory Data-Synthesis Pipeline (Collaborator, DeepWisdom) — I surveyed mainstream long-horizon agent benchmarks, distilling their data-synthesis schemes and test-task complexity, and produced a comparative analysis that guides how the pipeline should generate longer-horizon, more realistic trajectory data.
- IPV-Bench (first author) — the first systematic benchmark for image protection methods in image-to-video generation scenarios, unifying evaluation protocols and covering fidelity–effectiveness trade-offs, cross-model transfer, and robustness (arXiv:2603.26154, under review at AAAI 2027).
Education
- BEng in Artificial Intelligence, HKUST(GZ), 2023 – 2027 · CGPA 3.6+/4.3 (top 15%) · Machine Learning: A (top 15%) · Ethics, Privacy & Security in AI: A+
- Exchange Student, RWTH Aachen University, Aachen, Germany, Sep 2025 – Feb 2026 · Deep Learning (full marks)
Awards
- Third Prize, Huawei Ascend Operator Challenge (S8), 2026
- Merit Award, Undergraduate Research Project Competition, HKUST(GZ), 2024
- Dean’s List, HKUST(GZ), Spring 2025
Contact
- Email: xli683@connect.hkust-gz.edu.cn · Phone: +86 147 7150 6835
- GitHub: c-allergic
- CV: Download PDF · 中文版 PDF · CV page
I am currently seeking research and engineering internships in AI — feel free to reach out.
selected publications
-
IPV-Bench: Benchmarking Image Protection Methods under Diverse Image-to-Video Generation Scenarios2026Under review at AAAI 2027