Jason Qiang Sun
I am a Professor of Statistical Sciences, Computer Science, and Computer and Mathematical Sciences at the University of Toronto (UofT), where I lead the Stats + AI group. My long-term vision is to develop Artificial Intelligence (AI) that is efficient, trustworthy, personalized, and accessible to all. My current research focuses on efficient generative AI (GenAI), trustworthy AI, and advancing the foundations of next-generation statistics, driven by real-world challenges in technology, finance, and science. We actively collaborate with industry partners, so feel free to reach out.
Prior to my tenure at UofT, I was an associate research scholar at Princeton University. I earned my PhD from the University of North Carolina at Chapel Hill and my BS in SCGY from the University of Science and Technology of China. I am also recognized as a distinguished alumnus of UNC-CH.
In addition to my faculty role, I serve as an associate editor for Electronic Journal of Statistics and as an area chair for several ML conferences such as ICLR, COLT, AISTATS, and UAI. I also serve on the executive committee of BAAI young scientist club.
I may be slow to respond to emails. You can find herein my perspectives on the future of the statistics discipline and possible future research directions. Fundamentally, I champion problem- and data-centric approaches in statistics and AI, striving to achieve tangible advancements that benefit society and advance humanity.
If you are looking for information on recommendation letters, please see here.
Google Scholar | Research Gate | Github | 知乎 | X
Let's talk Stats + AI: A@B with A=qiang.sun B=utoronto.ca (academic, on leave),
C@D with C=qsunstats D=gmail.com (non-academic).
Address: Hydro 9042, 700 University Ave, M5G 1Z5 (UTSG),
IC 478, 1095 Military Trail, M1C 5J9 (UTSC).
Open Opportunities
We work on generative AI, trustworthy AI, and the foundation of next-generation statistics. We are currently recruiting top talents including students, interns, and visitors. If you are looking for such opportunities, please check here, complete this google form, and send me an email.
EverAI
We are building the EverAI open-door community for everyone, and you are welcome to join. Our mission is to foster a vibrant, inclusive community where everyone passionate about AI can come together to shape the future of AI.
Our first major event is the AI and Technology Summit at Toronto in January, 2025.
Recent News
[04/2025] Our paper on Markov-Bernstein inequalities was accepted in Annales de l'Institut Henri Poincaré.
[04/2025] Our paper on Unraveling when and why couriers wait is a recent top paper at SSRN.
[04/2025] Our paper PhysGame was accepted to the second workshop on computer vision for videogames (CV2) 2025.
[03/2025] Github page for our project on Lifelong learning with task-specific adaptation was online.
[03/2025] A recent university media coverage about our joint project: Drawing on AI and other data sciences to design next-gen joint replacements.
[02/2025] BAAI young scientist club annual conference.
[02/2025] Our paper Resist convex clustering: How does fusion penalty enhance robustness? was accepted to Electronic Journal of Statistics.
[01/2025] Two papers were accepted to ICLR 2025 and Bioinformatics, respectively.
[01/2025] AI and Technology Summit, Toronto.
[12/2024] Our outreach paper "Statistics and AI: A fireside conversation", which accompanies the Fireside chat on Stats and AI held on 03/17 , was accepted to Harvard Data Science Review.
[12/2024] Our PhysGame (LLM for games) benchmark and leaderboard are online.
[11/2024] Our paper Ensemble linear interpolators: The role of ensembling was accepted to SIAM Journal on Mathematics of Data Science. Congratulations to Mingqi! Bagging achieves a stabler performance in weak signal-to-noise ratio regimes while being consistent in large sample regimes. I refer to this ability as "algorithmic adaptivity". This notion of adaptivity holds the promise to explain "why do some algorithms always outperform other seemingly optimal algorithms?"
[11/2024] UNC 75 years anniversary conference and celebration.
[10/2024] Public online talk on Making AI trustworthy using statistics with over 300 audiences. A recording is available here.
[10/2024] Our recent paper on Decentralized online Riemannian optimization with dynamic environments was posted to arXiv.
[09/2024] Our recent paper on human nasal microbiome was accepted at Genome Biology.
[09/2024] Talk on making AI trustworthy at Lagrange research center, Paris (cancelled due to visa issues).
[09/2024] Seminar talk on "Making AI trustworthy" at IDS, HKU.
[08/2024] Plenary talk on trustworthy AI at ICT, Hangzhou.
[06/2024] Check our recent paper on Understanding large initialization accepted to UAI 2024.
[05/2024] Check our recent paper on Low rank matrix recovery under heavy-tailed data published at Bernoulli.
[05/2024] Plenary talk on Stats + Trustworthy AI @ STATSTRO 2024.
[04/2024] Hengchao's paper received the 2024 IMS Hannan Graduate Student Travel Award. Congratulations!
[03/2024] Our paper Variance-aware robust reinforcement learning with linear function approximation under heavy-tailed awards was accepted to TMLR.
[03/2024] Buxin's paper on The exact risks of reference panel-based regularized estimators received the 2024 Statistical Learning and Data Science (SLDS) Student Travel Award. Congratulations!
[03/2024] The Fireside chat on Stats and AI held on 03/17 invited by Professor Xihong Lin with over 1000 audiences. Here is the summary.
[03/2024] Our paper Quadratic matrix factorization with applications to manifold learning was accepted to TPAMI.
[01/2024] Our paper The exact risks of reference panel-based regularized estimators was posted on arXiv.
[01/2024] 1W-MINDS seminar (moved to March). Here is my recorded talk.
[10/2023] New version of the paper Do we need to estimate the variance in robust mean estimation? (Self-tuned robust mean estimators) was posted. The objective function proposed in this paper was referred to as the Sun-Huber objective by later works; see Holland (2023).
[10/2023] AI Talks @ MBZUAI.
[10/2023] Brown Bag Seminar.
[08/2023] Our paper Ensemble linear interpolators: The role of ensembling is now avaialble at arXiv. Bagging achieves a stabler performance in the weak signal-to-noise ratio regime while being consistent in strong signal-to-noise ratio regime. I refer to this ability as "algorithmic adaptivity". This notion of adaptivity holds the promise to explain "Why do some algorithms always outperform other seemingly optimal algorithms?" and serves a possible route to reliable/trustwothy AI.
[07/2023] ICML 2023.
[06/2023] Nuffield Econometrics Seminar at Oxford.
[05/2023] New paper on "Fast global convergence of gradient descent for low-rank matrix approximation" posted.
[05/2023] New version of the paper on "Self-tuned robust mean estimators (Do we need to estimate the variance in robust mean estimation?)". The objective function proposed in this paper was referred to as the Sun-Huber objective by later works; see Holland (2023).
[05/2023] Industry tech talk.
[05/2023] New paper on "Directional diffusion models".
[05/2023] New paper with Shuoguang on "Online generalized sparse regression: How does overparametrization help?".
[04/2023] The "Sketched ridgeless linear regression: The role of downsampling" paper has been accepted to ICML 2023. Congratulations to Xin, Yicheng, and Siyue!
[03/2023] Check our recent work on variance-aware robust reinforcement learnig, [arXiv]. Xiang is the first to come up with robust bandit and RL algorithms with tight variance-aware (instance-dependent) regrets!
[03/2023] Brown Bag Seminar.
[02/2023] Check our recent work on the statistical roles of downsampling: "Sketched ridgeless linear regression: The role of downsampling". Xin, Yicheng, and Siyue view downsampling as a dual of overparametrization, which then motivates new methods and theories!
[01/2023] I am serving as ACs for UAI 2023 and COLT 2023, and on the PC of KDD 2023 - ADS. Please consider submitting your best work at UAI 2023 and COLT 2023.
[10/2022] We are organizing an online reading seminar on (Reinforcement) Learning + X. Send me a message if interested.
[08/2022] I am serving as an AC for AISTATS 2023. Please consider submitting your best work at AISTATS 2023.
[07/2022] Together with Xingdong Feng, Chengchun Shi, Fan Zhou and Hongtu Zhu, we are organizing an online conference on ``From Statistics to Artificial Intelligence: Reinforcement Learning" at Shanghai this summer from 07/09 - 07/12. Thanks to SUFE, ARL and Mingshi for the support! Here is our conference flyer in Chinese and our conference flyer in English.
[04/2022] Talk at UNC-CH for the James E. Grizzle Distinguished Alumni Award from UNC-CH.
[04/2022] Congratulations to Dylan who successfully defended his thesis!
[03/2022] We are organizing an online reading seminar on Reinforcement Learning + X. Send me a message if interested.
[02/2022] New paper accepted to JASA.
[01/2022] I am serving on the PC for COLT 2022. Please consider submitting your best work at COLT 2022.
[12/2021] New paper accepted to Information and Inference.
[10/2021] I am happy to receive the James E. Grizzle Distinguished Alumni Award from UNC-CH.
[08/2021] New paper accepted to Integrating Materials and Manufacturing Innovation.
[07/2021] I am happy to serve as an area chair for AISTATS 2022. Please consider submitting your best work at http://aistats.org/aistats2022/.
[06/2021] New papers accepted to Electronic Journal of Statistics and Journal of Manufacturing Process.
[01/2021] I am happy to serve as a PC member for COLT 2021.
[11/2020] Congratulations to Dylan for passing the comprehensive exam.
[08/2020] New paper accepted to Journal of Machine Learning Research: Hoeffding's inequality for general Markov chains and its applications to statistical learning.
[06/2020] New paper accepted to Journal of Econometrics: Bayesian factor-adjusted sparse regression.
[06/2020] Congratulations to Anna Little for accepting an offer to join University of Utah as an assistant professor in the Department of Mathematics.
[05/2020] Congratulations to Qi Wang for accepting an offer to join Lanzhou University as an associate professor in the School of Life Sciences.
[05/2020] Congratulations to Rui Mao for receiving the prestigious IMS Hannan Graduate Student Award.
[04/2020] New paper accepted to JASA.
[03/2020] We are happy to receive the New Frontiers in Research Fund on machine learning for inverse material design.
[03/2020] New paper accepted to Nature Communications: Metagenome-wide association of gut microbiome features for schizophrenia.
[01/2020] New paper posted: Bayesian high-dimensional linear regression with generic spike-and-slab priors.
[01/2020] Read this nice commentary article about our recent paper on "Adaptive Huber regression": https://eranraviv.com/adaptive-huber-regression/.
[10/2019] Welcome to Mohamad Elmasri, who joins us as a postdoctoral fellow from MILA.
[09/2019] Welcome to Yicheng Zeng, who joins us as a postdoctoral fellow from HKBU.
[07/2019] New papers were accepted to Molecular Psychiatry and FEBS Letters.
[07/2019] New paper posted to arXiv: Robust convex clustering: How does fusion penalty enhance robustness?
[05/2019] Invited talk at UCSD.
[04/2019] Research day: http://www.fields.utoronto.ca/activities/18-19/stats-research-day.
[04/2019] Invited talks at UW Seattle, MSU, UIUC, UW Madison.
[03/2019] New paper accepted to Statistical Science: User-friendly covariance estimation for heavy-tailed distributions.
[03/2019] New paper posted on arXiv: Bayesian factor-adjusted sparse regression.
Seminars
"The weak and ignorance is not a barrier to survive, arrogance is." - <Three body problem> Liu CX
Learning on graphs and geometry | Valence labs
Machine leanring and mean field games
Stanford Causal Inference Seminar