Qi Huang 黄骑

M.Sc. student in Computational and Applied Mathematics, The University of Chicago

Qi Huang sitting on mossy stone steps in a forest

My research lies at the intersection of causal inference and machine learning for science. I am fortunate to work with Prof. Xinran Li at the University of Chicago, Prof. Zhuang Liu at Princeton University, and Prof. Xiao Luo at UW–Madison.

Before Chicago, I received my B.Sc. in Mathematics from the University of Liverpool with First Class Honours, ranked top 1% in my major. Outside of research, I’m into powerlifting 🏋️ and road trips 🚗.

Research

Causal inference

Sensitivity analysis for matched observational studies

Nov 2025 – present

Department of Statistics, University of Chicago

Mentor: Prof. Xinran Li

A studentized sensitivity analysis for Neyman’s weak null in matched observational studies.

Details
  • Anchors the within-set assignment law at fitted propensity odds rather than the uniform law.
  • The resulting Horvitz–Thompson lower envelope is mean-controlled and pointwise sharp, with asymptotically valid studentized tests; it extends in closed form to full matching and is applied to a right heart catheterization study.

Machine learning for science

Scientific world models & LLM agents

Mar 2026 – present

Zhuang Lab, Department of Computer Science, Princeton University

Mentor: Prof. Zhuang Liu

A scientific world model that predicts the outcomes of candidate research actions (research state + action → predicted outcome) to guide autonomous experiment planning with agents and vision-language models.

Details
  • Mechanism-aware predictors and a test-time selective-adaptation router for out-of-distribution generalization across unseen action families.
  • Reduced planner regret by ~80% in held-out replay evaluations.

Computational social science

Sep 2025 – Jul 2026

Knowledge Lab, University of Chicago

Mentor: Prof. James A. Evans

Led the mathematical modeling for a large-scale study of how AI breakthroughs concentrate scientific attention and erode intellectual diversity.

Details
  • Draws on 7.18M papers, 73K peer reviews and 3.7M scientists across 36 years.
  • Co-authored the resulting manuscript, submitted to Science.

Audio-visual test-time adaptation

Sep 2025 – Jun 2026

Department of Statistics, University of Wisconsin–Madison

Mentor: Prof. Xiao Luo

An audio-visual test-time adaptation framework for severe distribution shift that gates each modality with a binary accept/reject decision rather than softmax-style weighting.

Details
  • Learned attention maps with entropy-adaptive token dropout, curriculum annealing and class-specific thresholds reduce overfitting to noisy pseudo-labels.

Publications & Preprints

  1. Stronger Artificial Intelligence Breakthroughs Hasten Their Own Exhaustion

    ScienceSubmitted

    Beichen Lu, Honglin Bao, Qi Huang, Qingguo Meng, James A. Evans

    Science

  2. QEGF: A Quantum-Inspired Geometric and Energy-Based Fusion Framework for Multimodal Sentiment Analysis

    TPAMIMajor revision

    Yuxuan Hu, Yushan Pan, Minghui Zhu, Xiaojiang He, Qi Huang, Hao Wang, Yanlong Cao, Zhijie Xu

    IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)

  3. A Comprehensive Graph Pooling Benchmark: Effectiveness, Robustness and Generalizability

    TKDDMajor revision

    Pengyun Wang, Junyu Luo, Yanxin Shen, Ming Zhang, Shaoen Qin, Hanwen Xing, Siyu Heng, Qi Huang, Xiao Luo

    ACM Transactions on Knowledge Discovery from Data (TKDD)

  4. PVI-DPO: Perceptual Video Inpainting via Informative-Safe Preference Mining and Region-Aware Direct Preference Optimization

    AAAI 2027Submitted

    Jia Huang, Hongyou Zhou, Qi Huang, Hongbin Liu, Peng Zhang, Zhengqi Sun, Lumin Xu

    AAAI Conference on Artificial Intelligence (AAAI) 2027

News

  1. 🚀 Started research on scientific world models & LLM agents at Prof. Zhuang Liu’s lab @ Princeton University!

  2. 📊 Started working on causal inference for observational studies with Prof. Xinran Li @ UChicago Statistics.

  3. 🎉 Joined the University of Chicago as an M.Sc. student in Computational and Applied Mathematics!

  4. 🎧 Started audio-visual test-time adaptation research with Prof. Xiao Luo @ UW–Madison.

  5. 🎙️ Joined iFLYTEK as a machine learning research intern.

Background

Education

  1. The University of Chicago

    M.Sc. in Computational and Applied Mathematics

    Sep 2025 – Jun 2027Chicago, US

  2. The University of Liverpool

    B.Sc. in Mathematics, First Class Honours (top 1%)

    Sep 2023 – Jun 2025Liverpool, UK

  3. Xi’an Jiaotong-Liverpool University

    B.Sc. in Applied Mathematics

    Sep 2021 – Jul 2023Suzhou, China

Industry

  1. iFLYTEK

    Machine Learning Research Intern

    Trained Transformer baselines on text (BLEU 27.6 to 29.9) and Conformer baselines on speech (WER 10.3% to 8.9%).

    Jun 2025 – Sep 2025Hefei, China

  2. Alibaba Group

    Data Analyst Intern

    Consumer-data analysis with AARRR metrics and RFM-based segmentation for retention analysis and precision marketing.

    Jul 2024 – Aug 2024Hangzhou, China

Service

  1. Forecast@ICML 2026 Workshop

    Reviewer

    Forecasting as a New Frontier of Intelligence