Zihua She

Zihua She

Purdue University

📝 Publications

Submitted and Under Review

  • What Zero Detections Certify: Statistical Safety Cases for Model-Aware Monitoring. Zihua She, Xiao Wang. Submitted. GitHub repository, GitHub Page.

  • LLM Benchmark Means can be blind to the Tail. Zihua She, Xiao Wang. Submitted. GitHub repository, GitHub Page.

  • Your FID cannot see the Tail: Auditing and Reparing the Rare-Event Calibration of Large Generative Models. Zihua She, Xiao Wang. Submitted. GitHub repository, GitHub Page.

  • Optimal Estimation for Flow Matching via Coreset Induced Coupling. Xiao Wang, Zihua She, Jianxi Su. Submitted to The Annals of Statistics.

Preprint

  • Which Self-Improvements Should We Trust? Certified Promotion When Agents Reuse Their Benchmark. Xiaojing Sun, Yuhan Zeng, Zihua She, Xiao Wang. Submitted. arXiv.

  • Coreset-Induced Conditional Velocity Flow Matching. Zihua She, Jianxi Su, Xiao Wang. Accepted to The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS 2026). arXiv, GitHub repository, GitHub Page.

Technical Report

  • AI-Based Synthetic Medical Claims Data Generation: A Practical Guide for Actuaries. Shea Parkes, Zihua She, Jianxi Su, Xiao Wang. The Society of Actuaries. SOA report, GitHub repository.

Working Paper

  • Score and Rank Inference for Large Language Model. Zihua She, Xiao Wang. Manuscript in preparation.

  • Half A Tweedie Step. Zihua She, Xiao Wang. Manuscript in preparation.

  • Coreset-Based Joint Source-Coupling Design for Flow Matching. Zihua She, Jianxi Su, Xiao Wang. Manuscript in preparation.

Book Chapter

  • BIG DATA IN FINANCE: Big Data and Industrial Chain Analysis. ISBN: 978-7-5654-4751-8. In Chinese 《金融大数据：大数据与产业链分析》. 陈玉罡、黎江、罗佳璟、佘梓华.

🛠️ Software

  • TeX Proof-Check Kit: Checking the Proofs in LaTeX Papers with Lean 4 and AI Agents. Zihua She, Xiao Wang. Open-source software, v1.3, 2026. GitHub repository.

👨🏻‍🏫 Talks & Posters

Coreset-Based Joint Source-Coupling Design for Flow Matching

  • IMSI Workshop, Statistical Foundations of Generative Modeling

  • The University of Chicago, Chicago, October 5 - 9, 2026.

Last updated: October 10, 2026