Research Profile
Machine Learning · Probabilistic AI
Zhuo Sun
Tenure-track Assistant Professor
School of Statistics and Data Science, SUFE
I develop efficient, reliable, and scalable learning algorithms, with a focus on reinforcement learning, foundation models, probabilistic machine learning, and efficient inference.

Visiting Researcher Imperial College London
Academic Service Area Chair / PC ICML · ICLR · NeurIPS · AISTATS · UAI
Recognition Best Paper · Oral · Spotlight
About
I received my Ph.D. from University College London, supervised by Prof. François-Xavier Briol and Prof. Jing-Hao Xue. I completed my M.Sc. at the University of Oxford, supervised by Prof. George Deligiannidis and mentored by Prof. Gesine Reinert.
Reinforcement Learning Foundation Models Probabilistic ML Efficient Inference
News
- Sep 2026 — 3 papers are accepted to NeurIPS 2026!
- Jun 2026 — Interdomain Attention: Beyond Token-Level Key-Value Memory is released on arXiv and selected as a Spotlight Paper at the ICML 2026 Workshop on Foundations of Deep Generative Models.
- Apr 2026 — 1 paper is accepted to ICML 2026!
- Feb 2026 — Information Shapes Koopman Representation is selected as an Oral Paper at ICLR 2026!
- Jan 2026 — 3 papers are accepted to ICLR 2026!
Publications & Preprints
* equal contribution; ✉ corresponding author
arxiv
2026
(2026+). SR-OPSD: Self-Referenced On-Policy Self-Distillation arxiv. arXiv
arxiv
2026
arxiv
2026
arxiv
2026
NeurIPS
2026
(2026). Fisher Decorator: Refining Flow Policy via A Local Transport Map In The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026. arXiv
NeurIPS
2026
(2026). Outlier-Robust Diffusion Posterior Sampling for Bayesian Inverse Problems In The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026. arXiv
NeurIPS
2026
(2026). Random-Projection Tree Stein Variational Gradient Descent In The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026. arXiv
ICML
2026
(2026). How Does the Lagrangian Guide Safe Reinforcement Learning through Diffusion Models? In Proceedings of the 43rd International Conference on Machine Learning (ICML 2026). arXiv
ICLR
2026
(2026). Multilevel Control Functional. In International Conference on Learning Representations (ICLR 2026). arXiv With Score 8,8,8 (Rank 2nd over 19000 submissions)
ICLR
2026
(2026). Information Shapes Koopman Representation. In International Conference on Learning Representations (ICLR 2026). arXiv Selected for Oral Presentation (top 1.18%)
ICLR
2026
(2026). From Embedding to Control: Representations for Stochastic Multi-Object Systems. In International Conference on Learning Representations (ICLR 2026). arXiv
UAI
2023
(2023). Meta-learning Control Variates: Variance Reduction with Limited Data. In Proceedings of the 39th Conference on Uncertainty in Artificial Intelligence (UAI 2023). Selected for Oral Presentation (top 3%)
ICML'W
2023
(2023). Multilevel Control Functional (Short Version). In ICML 2023 SPIGM.
ICML
2023
(2023). Vector-valued Control Variates. In Proceedings of the 40th International Conference on Machine Learning (ICML 2023). ASA SBSS Best Student Paper (2022)
AISTATS
2021
(2021). Amortized Bayesian Prototype Meta-learning: A new probabilistic meta-learning approach to few-shot image classification. In Proceedings of the 24th International Conference on Artificial Intelligence and Statistics(AISTATS 2023).
Neurocomputing
2021
(2021). A Concise Review of Recent Few-shot Meta-learning Methods. Neurocomputing. arXiv
IEEE TIP
2020
(2020). Bi-Similarity Network for Fine-grained Few-shot Image Classification. IEEE Transactions on Image Processing. arXiv

