CV

Hongkun Yang

Undergraduate researcher in responsible AI

hongkunyang16@gmail.com
(+86) 152-7065-8828
Qingdao, , CN

Summary

Undergraduate student in Mathematics and Applied Mathematics at Ocean University of China. Research interests include robust and reliable AI, vision-language models, knowledge distillation, and responsible AI for real-world systems.

Education

  • B.S. in Mathematics and Applied Mathematics
    Present
    Ocean University of China (OUC), Qingdao, China
    GPA: 86/100 (3.5/4.0)
    Courses: Data Structures (99/100), Mathematical Analysis (98/100), Higher Algebra (97/100)

Work Experience

  • Research Assistant
    Jun. 2026 - Present
    Rice University, Houston, TX, USA
    Adviser: Prof. Wei Qiu
    • Developed PerDFKD, a source-data-free knowledge-distillation framework that reconstructs personalized ECG teachers and transfers their knowledge to lightweight student models.
  • Research Assistant
    Jul. 2025 - Sep. 2025
    Nanyang Technological University, Singapore
    Adviser: Prof. Wei Dong
    • Proposed a non-transferable-learning framework for vision-language models to prevent unauthorized fine-tuning and intellectual-property exploitation while preserving utility on authorized tasks.

Publications

  • DeAR: Fine-Grained VLM Adaptation by Decomposing Attention Head Roles
    2026
    CVPR 2026
    Yiming Ma*, Hongkun Yang*, Lionel Z. Wang, et al. Accepted to CVPR 2026. Equal contribution. Refined the core framework, led most experiments, and co-authored the manuscript.
  • PerDFKD: Source-Data-Free Knowledge Distillation for Efficient Personalized ECG Adaptation
    2026
    Under Review at KDD 2027
    Hongkun Yang, Yilin Song, Wei Qiu. Designed the complete PerDFKD algorithm, implemented all experiments, and wrote the manuscript.
  • AppellateGen: A Benchmark for Appellate Legal Judgment Generation
    2026
    Under Review
    Hongkun Yang, Lionel Z. Wang, Wei Fan, et al. Built the benchmark, developed the SOP-based multi-agent system, and led evaluations and writing.
  • NüshuVoice: Reviving the Voice of Endangered Nüshu with Pitch-Aware Text-to-Speech
    2026
    Under Review
    Hongkun Yang*, Xinhui Yi*, Xiyan Zhao*, Yibo Meng*, et al. Equal contribution. Built the first Nüshu TTS benchmark and contributed to the pitch-aware Nüshu-PitchVITS framework.

Teaching

  • Algorithms and Data Structures
    2025
    Ocean University of China (OUC), Qingdao, China
    Role: Teaching Assistant (Sep. 2025–Jan. 2026)
    Conducted weekly office hours and Q&A sessions, provided debugging guidance, and evaluated programming assignments and examinations.