Jiyuan Tan
PhD Student, Management Science and Engineering · Stanford University
Causal inference · machine learning · trustworthy automation for data science
About
I am a third-year PhD student in Management Science and Engineering at Stanford University, co-advised by Prof. Vasilis Syrgkanis and Prof. Jose Blanchet.
My research asks how far we can automate causal inference — from the statistical estimator to the AI agent — without giving up the mathematical guarantees, transparency, and domain sensitivity that scientific and policy applications require. It sits at the boundary of statistics, econometrics, machine learning, and human-centered AI.
Before Stanford, I studied mathematics at Fudan University. There I worked with Prof. Yinyu Ye on SOLNP+, a derivative-free solver for general nonlinear optimization, and with Prof. Zhaoran Wang and Prof. Zhuoran Yang on offline learning in zero-sum Markov games.
Research
-
Partial identification
When data and assumptions cannot pin down a single causal effect, they often still imply informative bounds. I build methods that compute those bounds with statistical guarantees.
-
Automating Causal Inference
Causal estimation is full of choices that today demand an expert: regularization, tuning, estimator selection. I make those choices data-driven while keeping the theory intact.
-
Causal reasoning in AI
Can we trust an AI system's causal reasoning? I build benchmarks that test it, methods that audit a model's internal computation, and Lean-based formal verification for causal results.
News
- Oct 2026Released PaperFormalizer, an author-in-the-loop tool that turns a mathematical paper into a Lean-verified version: it checks each result against the paper's own proof, proposes revisions the author decides on, and proves the statements in Lean 4. The code is open, and formalized papers are browsable on the website.
- Jul 2026New preprint: CausalSmith: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference, with Vasilis Syrgkanis. Machine-verified working papers and the Causalean library are browsable on the project website.
- May 2026New preprint: Bucketing the Good Apples: A Method for Diagnosing and Improving Causal Abstraction, with Puyin Li, Ahmad Jabbar, Thomas Icard, and Atticus Geiger.
- Apr 2026New preprint: Partial Identification of Policy-Relevant Treatment Effects with Instrumental Variables via Optimal Transport, with Vasilis Syrgkanis and Jose Blanchet.
- Mar 2026New preprint: Adaptive Estimation and Inference in Conditional Moment Models via the Discrepancy Principle, with Vasilis Syrgkanis.
- Feb 2026New preprint: CausalReasoningBenchmark: A Real-World Benchmark for Disentangled Evaluation of Causal Identification and Estimation, with Ayush Sawarni and Vasilis Syrgkanis.
- Oct 2025Talk on Consistency of Neural Causal Partial Identification at the INFORMS Annual Meeting.
- Sep 2025Estimation of Treatment Effects in Extreme and Unobserved Data was accepted at NeurIPS 2025.
- Jun 2025Started a summer internship as a Data Scientist at LinkedIn, working on long-term treatment effects with debiased machine learning.
Contact
The fastest way to reach me is jiyuantan@stanford.edu. I am always glad to chat about everything.