About Me

I am pursuing a bachelor’s degree at The Hong Kong University of Science and Technology (HKUST), double majoring in Mathematics and Computer Science. I am currently working on Conformal Prediction and Selective Inference, where I am fortunate to be advised by Prof. Ying Jin at the Wharton School of the University of Pennsylvania. Additionally, I am exploring Conformal Prediction with Latent Structure under the supervision of Prof. Xinzhou Guo in the Mathematics Department at HKUST. Before that, I was a research assistant working on Cognitive Science and Intention Understanding at the KnowComp Group at HKUST, where I have been fortunate to be advised by Prof. Yangqiu Song. I visited the School of Mathematics at ETH Zurich as a university exchange student during the spring of 2025.

Under Review

[1]   Multi-Distribution Robust Conformal Prediction [Code] [Paper]

[2]   SessionIntentBench: Multi-task Inter-Session Intention-Shift Modeling Benchmark for E-commerce Customer Behavior Understanding [Code] [Paper]

Research Experience

Sep2025-Present   Capstone Project at the Mathematics Department, HKUST, adviced by Prof. Xinzhou Guo

• Conformal prediction with latent structure, leveraging effective latent cluster modeling at both the covariate and residual levels; has already demonstrated strong empirical performance compared to existing methods such as PCP (Posterior Conformal Prediction). [preliminary results]

May2025-Present   Research Assistant at Wharton, adviced by Prof. Ying Jin

• Developed a novel conformal prediction method addressing both classification and regression problems in a multi-source environment, where data are drawn from heterogeneous distributions. Established conformal set with high efficiency and tight worst-case coverage under both marginal and conditional settings.

Apr2024-Feb2025   Research Assistant at KnowComp Group, adviced by Prof. Yangqiu Song

• Proposed new intention modeling frameworks for underexplored session interactions:
  • developed a paradigm that is easy to deploy, automate and scale
• Implemented the whole benchmarking / dataset generation pipeline:
  • data cleaning, preprocessing
  • prompt engineering: Zero-shot, Few-shots, COT
  • L(V)LM generations: feature extraction, aligned generation, aligned verification
  • data labeling stage: annotation protocol writing, testing and evaluation
  • model evaluation: LLM, LVLM / open-source models (mostly Transformer models), proprietary ​API models
  • model finetuning: LLM, LVLM finetuning via Llama-Factory
  • structural codes: coding the logics and structures that enable above functionalities

Oct2024-Dec2024   Research Project adviced by Prof. Can Yang

Feb2024-May2024   Research Project adviced by Prof. Qifeng Chen

Industry Experience

Jan2024-Jan2024   Research Analyst Intern at Jihai Investment (Hedge Fund)

Dec2023-Jan2024   Investment Banking Intern at Zhe Shang Securities

Jun2023-Aug2023   Software Developer Intern at Hundsun (world top 50 Fintech)

Projects

[1]   Gaussian Process Boosting Theories and Simulation Studies

[2]   Deep Convolutional GAN and CycleGAN for Image Processing and Generation [Code]

Awards

Aug2024   China Construction Bank Asia Scholarships (top 3 first prize winning awardee)

May2024   The S.S. Chern Class Scholarship (top 20 math student researchers)

Apr2024   Chiaphua Industries Limited Scholarships (top 1% tier invitation and final awardee)

Oct2023   University’s Scholarship Scheme for Continuing UG Student (top 1% tier)

Jun2022   National College Entrance Examination (GaoKao) (top 0.5% among 360,000 participants)

Contact

Reach me at yyangfd [at] connect [dot] ust [dot] hk for potential collaborations or visit my GitHub / LinkedIn home page for more information.