My research focused on probabilistic ML and generative modelling, particularly on the scalable (approximate) sampling methods (VI/MCMC/SMC/etc.) with guarantees.

I remain active in the research community and am open to academic collaborations. Some problems that I find intriguing lately include:

  • Tabular foundation model for time-series prediction and causal impact inference;
  • Statistical methods for highly-skewed time-series (e.g., modeling "heavy-tail" distributions where a small percentage of users drives the majority of revenue);
  • Inference-time control of generative models;
  • Application of Monte Carlo methods and flow-based generative modelling in rendering.

* denotes equal contribution.

Preprints and Workshops

Automated Discovery of Pairwise Interactions from Unstructured Data
Zuheng Xu*, Moksh Jain*, Ali Denton, Shawn Whitfield, Aniket Didolkar, Berton Earnshaw, Jason Hartford
[arXiv]
Score-Based Interaction Testing in Pairwise Experiments
Jana Osea*, Zuheng Xu*, Cian Eastwood, Jason Hartford
Causal Representation Learning Workshop @NeurIPS, 2024
[CRL@NeurIPS] [poster]

Publications

Asymptotically exact variational flows via involutive MCMC kernels
Zuheng Xu, Trevor Campbell
Advances in Neural Information Processing Systems, 2025 (Accepted)
[arXiv] [NeurIPS]
Tuning Sequential Monte Carlo Samplers via Greedy Incremental Divergence Minimization
Kyurae Kim*, Zuheng Xu*, Jacob R. Gardner, Trevor Campbell
International Conference on Machine Learning, 2025
[arXiv] [ICML]
Is Gibbs sampling faster than Hamiltonian Monte Carlo on GLMs?
Son Luu, Zuheng Xu, Nikola Surjanovic, Miguel Biron-Lattes, Trevor Campbell, Alexandre Bouchard-Côté
International Conference on Artificial Intelligence and Statistics, 2025
[arXiv] [AISTATS]
Propensity Score Alignment of Unpaired Multimodal Data
Johnny Xi, Jana Osea, Zuheng Xu, Jason Hartford
Advances in Neural Information Processing Systems, 2024
[arXiv] [NeurIPS]
Turning waste into wealth: leveraging low-quality samples for enhancing continuous conditional generative adversarial networks
Xin Ding, Yongwei Wang, Zuheng Xu
AAAI Conference on Artificial Intelligence, 2024
[arXiv] [AAAI]
Embracing the chaos: analysis and diagnosis of numerical instability in variational flows
Zuheng Xu, Trevor Campbell
Advances in Neural Information Processing Systems, 2023
[arXiv] [NeurIPS] [poster] [slides]
MixFlows: principled variational inference via mixed flows
Zuheng Xu, Naitong Chen, Trevor Campbell
International Conference on Machine Learning, 2023
[arXiv] [code] [ICML] [poster]
Bayesian inference via sparse Hamiltonian flows
Naitong Chen, Zuheng Xu, Trevor Campbell
Advances in Neural Information Processing Systems (oral), 2022
[arXiv] [code] [NeurIPS]
Distilling and transferring knowledge via cGAN-generated samples for image classification and regression
Xin Ding, Yongwei Wang, Zuheng Xu, Z. Jane Wang, William J. Welch
Expert Systems with Applications, 2023
[arXiv] [code] [ESA]
Continuous conditional generative adversarial networks: novel empirical losses and label input mechanisms
Xin Ding, Yongwei Wang, Zuheng Xu, Z. Jane Wang, William J. Welch
IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023
[arXiv] [code] [TPAMI]
The computational asymptotics of Gaussian variational inference and the Laplace approximation
Zuheng Xu, Trevor Campbell
Statistics and Computing 32(63), 2022
[arXiv] [code] [Statistics and Computing] [Msc. Thesis] [slides]
CcGAN: continuous conditional generative adversarial networks for image generation
Xin Ding, Yongwei Wang, Zuheng Xu, William J. Welch, Z. Jane Wang
International Conference on Learning Representations, 2021
[ICLR] [code]
Bayesian pseudocoresets
Dionysis Manousakas, Zuheng Xu, Cecilia Mascolo, Trevor Campbell
Advances in Neural Information Processing Systems, 2020
[NeurIPS] [code]
Student submission patterns in online homework and relationships to learning outcomes: a pilot study
Gianni Co, Zuheng Xu, Giorgio Sgarbi, Siqi Cheng, Ziqi Xu, Agnes d'Entremont, Juan Abelló
Proceedings of the Canadian Engineering Education Association Conference, 2019
[CEEA]