Yu Gui

National Studies on Air Pollution and Health

I am a Postdoctoral Researcher in the Department of Statistics and Data Science at the Wharton School, University of Pennsylvania, working with Professors Dylan Small and Zhimei Ren.

I obtained my PhD in Statistics at the University of Chicago, where I was fortunate to be advised by Professors Rina Foygel Barber and Cong Ma.

Prior to my PhD, I graduated from School of the Gifted Young at University of Science and Technology of China and was a student research intern advised by Professor Jun S Liu at Harvard.

I am broadly interested in human-in-the-loop statistical inference and statistical learning with multi-modalities, motivated by problems that arise from and, in turn, inform real-world applications.

I am on the 2026-2027 academic job market.

research interests

Reliable, robust, and adaptive statistical inference
- Adaptive and robust statistical methods in causal inference, addressing challenges of heterogeneity, multiplicity, unmeasured confounding, and selection bias (GSR26).
- Distribution-free uncertainty quantification for black-box models, including reliable and human-in-the-loop deployment of foundation model outputs with statistical guarantees (GJR24, GJNR25).
- Robust inference in the presence of distribution shift, missingness, selection bias, and censoring (GBM23, GBM24, GHRB24).
Statistical machine learning with multi-modal data
- Theoretical foundations of modern (self-supervised) representation learning techniques (GMM25a, GMZ23)
- Efficient methods for disentangling shared and modality-specific features in spatial single-cell multi-omics applications (GMM25b).

news

Jun 27, 2026	Our paper Distributionally robust risk evaluation with an isotonic constraint has been accepted to Information and Inference: A Journal of the IMA. This paper offers an approach that utilizes side information (e.g. shape constraints) to balance the misspecification of sample reweighting and the over-pessimism of distributionally robust learning!
May 11, 2026	New preprint Adaptive discovery of effect modification in matched observational studies: a finite-sample valid procedure to identify covariate-interpretable subgroups with exact subgroup-level FDR control, robust to unmeasured confounding via sensitivity models and powered by multiple matched controls.
Sep 28, 2025	Multi-modal contrastive learning adapts to intrinsic dimensions has been accepted to NeurIPS 2025. We present a theoretical analysis of CLIP, showing how temperature optimization enables adaptation to the intrinsic dimension of shared features in multi-modal data (poster). A more recent work proposes IndiSeek which learns modality-specific features that are independent of shared features.
Sep 17, 2025	I’m honored to be awarded the IMS Lawrence D. Brown Ph.D. Award!
Apr 02, 2025	Our paper Conformal prediction: a data perspective has been accepted to ACM Computing Surveys.
Sep 26, 2024	Our paper Conformal alignment has been accepted to NeurIPS 2024! This paper guarantees the safe and reliable deployment of foundation model outputs (poster). A more recent paper ACS on interactive conformal selection.
Jun 30, 2024	I’m happy to be awarded the William Rainey Harper Dissertation Fellowship!
Dec 31, 2023	Presented Conformalized matrix completion at NeurIPS 2023 and ICSDS 2023 (poster).