Publications

Also on Google Scholar. An asterisk (*) marks equal contribution or alphabetical author order.

Preprints

• Preprint
  2026
  Bucketing the Good Apples: A Method for Diagnosing and Improving Causal Abstraction Puyin Li*, Jiyuan Tan*, Ahmad Jabbar, Thomas Icard, Atticus Geiger A diagnostic for causal abstraction in neural networks, used to localize and repair failures of interpretability probes on language models. arXiv
• Preprint
  2026
  Partial Identification of Policy-Relevant Treatment Effects with Instrumental Variables via Optimal Transport Jiyuan Tan*, Vasilis Syrgkanis, Jose Blanchet Casts partial identification with instrumental variables as an optimal-transport problem, giving a tractable estimator with finite-sample guarantees. arXiv
• Preprint
  2026
  Adaptive Estimation and Inference in Conditional Moment Models via the Discrepancy Principle Jiyuan Tan*, Vasilis Syrgkanis Data-driven regularization for conditional moment models, including nonparametric IV, with oracle inequalities and valid inference. arXiv
• Preprint
  2026
  CausalReasoningBenchmark: A Real-World Benchmark for Disentangled Evaluation of Causal Identification and Estimation Ayush Sawarni, Jiyuan Tan, Vasilis Syrgkanis A benchmark built from published empirical studies that separates causal identification from estimation, enabling fine-grained evaluation of language models and causal methods. arXiv

Journal publications

• ACM TOMS
  2024
  Algorithm 1053: SOLNP+: A Derivative-Free Solver for Constrained Nonlinear Optimization Dongdong Ge, Jinsong Liu, Tianhao Liu, Jiyuan Tan, Yinyu Ye* (alphabetical order) ACM Transactions on Mathematical Software, 50(4), Article 29, 1–24. An ANSI C solver for constrained nonlinear optimization, made robust to noisy function evaluations via implicit filtering and coordinate search. PaperarXivCode

Conference publications

• NeurIPS
  2025
  Estimation of Treatment Effects in Extreme and Unobserved Data Jiyuan Tan*, Jose Blanchet, Vasilis Syrgkanis An extreme-value estimator for treatment effects beyond the observed support of the data, with consistency guarantees and validation on synthetic and real data. arXiv
• NeurIPS
  2024
  Consistency of Neural Causal Partial Identification Jiyuan Tan, Jose Blanchet, Vasilis Syrgkanis A generative method for computing partial-identification bounds under structural causal constraints, with a consistency proof and an analysis of the model class's approximation power. arXiv
• UAI
  2024
  A Homogenization Approach for Gradient-Dominated Stochastic Optimization Jiyuan Tan*, Chenyu Xue*, Chuwen Zhang, Qi Deng, Dongdong Ge, Yinyu Ye A homogenized second-order method for gradient-dominated stochastic optimization, including policy optimization, with state-of-the-art sample-complexity guarantees. arXiv
• ICML
  2022
  Pessimistic Minimax Value Iteration: Provably Efficient Equilibrium Learning from Offline Datasets Han Zhong*, Wei Xiong*, Jiyuan Tan*, Liwei Wang, Tong Zhang, Zhaoran Wang, Zhuoran Yang Characterizes the data coverage needed to learn Nash equilibria from offline Markov-game data, with a pessimistic algorithm attaining minimax-optimal sample complexity. Spotlight at the ICLR 2022 Workshop on Gamification and Multiagent Solutions. arXiv

Talks

• 2025Consistency of Neural Causal Partial Identification — INFORMS Annual Meeting.
• 2024Consistency of Neural Causal Partial Identification — California Econometrics Conference (CEC).