Zihan Huang

Zihan Huang

I am currently a Master Student in Computer Science at UC, San Diego. I obtained my bachelor’s degree from Turing Class, Chu Kochen Honors College of Zhejiang University.

🔥 News

  • 2024.09:  🎉🎉 One paper accepted by Briefings in Bioinformatics (paper link)
  • 2024.06:  🎉🎉 Open-sourced my project AutoGeo (project page)
  • 2024.06:  🎉🎉 Graduated from Chu Kochen Honors College of Zhejiang University with B.E.

📖 Educations

  • 2024.09 - Present B.S. in Computer Science, University of California San Diego, CA, United States
  • 2020.09 - 2024.06 B.E. in Artificial Intelligence, Turing Class, Chu Kochen Honors College, Zhejiang University, Hangzhou, China

📝 Research and Work Experiences

Alibaba Tmall Meta Research Group, 2024.6-Present

Shanghai Institute for Advanced Study of Zhejiang University, 2024.1-2024.6, Supervisor : Prof. Jingyuan Chen

INeurIPS 2024 Track on Datasets and Benchmarks
sym

AutoGeo: Automating Geometric Image Dataset Creation for Enhanced Geometry Understanding, NeurIPS 2024 Datasets and Benchmarks Track (Under Review)

Zihan Huang, Tao Wu, Wang Lin, Shengyu Zhang, Jingyuan Chen, Fei Wu

  • We created a tool to automatically generate various images of geometry figures and their corresponding natural language descriptions.
  • We generate a datasets AutoGeo-100K and use it for multi-modal LLMs alignment training and verify the reliable of our datasets through this.

ZJU DCD Lab, 2022.2-2023.6, Supervisor : Prof. Fei Wu

BIB
sym

Global-local aware heterogeneous graph contrastive learning for multifaceted association prediction in miRNA-gene-disease networks

  • We proposed a contrastive pre-training methods of GNN to predict the correlation between miRNA-gene and diseases.
IEEE TMM
sym

Private Data protected Distillation model for image captioning via multi-teachers latent space feature aggregation

Mengze Li, Zihan Huang, Ziqi Tan, Zhengqing Fang, Zhou Zhao, Kun Kuang, Shengyu Zhang, Fei Wu

  • We created an framework based on differential privacy to filter privacy information in data while training captioning model.
  • The evaluation result of our framework PDD shows that we can add more privacy protection for models while keeping maintaining better model usability compared with other methods.
ACM-MM2023
sym

TITAN: Test-tIme Training via counterfactuAl generation and casual interventioN for text-video retrieval

Mengze Li, Zihan Huang, Han Wang, Jiaxu Miao, Wenyan Fan, Zhou Zhao, Shengyu Zhang, Wenming Tan, Fei Wu

  • We created an system with causal disentanglement to retrieve videos of different data distribution by natural language discription of their contents.
  • The result and case study shows of our systems can retrieve videos with differnent senarios robustly.
  • We provides a datasets with various videos from life and movie to better evaluate our system and for research in this area.

🎖 Honors and Awards

  • Project leader of National Student Research Training Program(SRTP) and rated as Outstanding, 2022-2023
  • Honorable Winner of Mathematical Contest in Modeling(MCM), work as team leader, 2022, COMAP
  • Academic Excellence Award 2020-2022, Chu Kochen Honors College of ZJU
  • Second-Class Scholarship for Elite Student in Basic Sciences, 2021-2022, ZJU
  • Serve as NeurIPS emergency Reveiwer, 2023
  • 2nd Prize of 3D Printing Competition, 2020, ZJU
  • 2nd Prize of Zhejiang College Student Physics Competition, 2021
  • Half Marathon 1h53m and Ranked 84/1000+, 2020, HangZhou, Zhejiang
  • 1st Prize of Search&Rescue of Asia-Pacific Robotics Championship(APRC), 2013
  • 1st Prize of Triathlon of Asia-Pacific Robotics Championship(APRC), 2013