DeepThink Lab

Dec 2026 Tutorial: Presenting a 3-hour tutorial at NeurIPS’26 (Sydney) on “Principles of Deep Representation Learning via Low-Dimensional Models”, together with Prof. Yi Ma and Prof. Peng Wang

Nov 2026 Minitutorial: Presenting a 2-hour minitutorial at the SIAM Conference on Mathematics of Data Science (MDS26) on “Diffusion Generative Models: From Fundamentals to Modern Developments”, together with Prof. Molei Tao (Georgia Tech)

Oct 2026 Conference: At the Asilomar Conference on Signals, Systems, and Computers (Oct 25–28, Pacific Grove, CA), co-chairing the session Generative Modeling for Inverse Problems with Prof. Yu Sun (Johns Hopkins), and presenting Compressible Dynamics in Deep Overparameterized Low-Rank Learning & Adaptation and ForcingDAS: Long-Sequence Data Assimilation with Diffusion Forcing

Sep 2026 Workshop: BIRS Workshop on Mathematical Foundations in Deep Learning and Generative AI

Aug 2026 Workshop: NUS & IMS Mathematical Foundations of AI Models

Jul 2026 Workshop: Organizing a workshop on Foundations of Deep Generative Models: Understanding Memorization, Generalization, and Reasoning at ICML’26!

Jul 2026 Summer Course: Taught a four-day introductory course at ESSAI 2026 (Jul 6–9, Vienna, Austria) on “Learning Deep Low-Dimensional Models from High-Dimensional Data: From Theory to Practice”, together with Dr. Sam Buchanan (UC Berkeley), Prof. Yi Ma (HKU), and Prof. Zhihui Zhu (Ohio State) (Lecture 1, Lecture 2, Lecture 3, Lecture 4)

Jun 2026 Talk: Midwest Machine Learning Symposium (MMLS)

Jun 2026 Workshop: ACM SIGMETRICS 2026 Workshop on Frontiers in Generative AI: Foundation and Algorithms

Jun 2026 Tutorial: ISIT 2026 (with Yuxin Chen, Yuting Wei, Liyue Shen): Harnessing Low Dimensionality in Diffusion Models: From Theory to Practice

Jun 2026 Mini-symposium: Optimization Foundations of Generative AI Models at SIAM OP’26

May 2026 Tutorial: CISA 2026 (with Liyue Shen): Deep Generative Models for Inverse Problems

May 2026 Tutorial: ICASSP 2026 (with Yuxin Chen, Yuting Wei, Liyue Shen): Harnessing Low Dimensionality in Diffusion Models: From Theory to Practice

Apr 2026 Workshop: Co-organizing 2nd DeLTa Workshop (Deep Generative Model in Machine Learning: Theory, Principle and Efficacy) at ICLR’26

Mar 2026 Invited Talk: Out-of-Distribution Generalization of In-Context Learning: A Low-Dimensional Subspace Perspective at SIAM UQ’26

Feb 2026 Invited Talk: Generalization of Diffusion Models Arises with a Balanced Representation Space at ITA Workshop

JAN 2026 Invited Talk (Video): Presenting our work at The One World Seminar Series on the Mathematics of Machine Learning on Understanding Generalization of Deep Generative Models based on Low-dimensional Structures

Oct 2025 Tutorial (Website): Presenting a full-day tutorial at ICCV’25 on “Learning Deep Low-Dimensional Models from High-Dimensional Data: From Theory to Practice“, together with Profs. Yi Ma, Liyue Shen, Atlas Wang, Zhihui Zhu , and Dr. Sam Buchanan

Sep 2025 Lectures: Giving 3 Lectures at IAISS'25: (Lecture I, Lecture II, Lecture III)

Jul 2025 Tutorial (Website): Presenting a 2.5-hour tutorial at ICML’25 on “Harnessing Low Dimensionality in Diffusion Models: From Theory to Practice” with Prof. Liyue Shen and Prof. Yuxin Chen

Oct 2026 Paper Acceptance: Our paper The Linear Geometry of Interpretable Tokens: Jailbreaking Attacks and Defenses for Unlearned Diffusion Models has been accepted at Transactions on Machine Learning Research (TMLR), with a J2C Certification (Project Page).

Oct 2026 Paper Release: 3 new preprints from our group, on diffusion sampling for PDE inverse problems, in-context learning in neural quantum states, and text embeddings for diffusion models

Sep 2026 Paper Acceptance: Our paper Emergent Low-Rank Training Dynamics in MLPs with Smooth Activations has been accepted at NeurIPS’26.

Sep 2026 Paper Acceptance: Our paper VoLo: A Physical Orchestrator for Open-Vocabulary Long-Horizon Manipulation has been accepted at CoRL’26 (Project Page).

Jun 2026 Paper Acceptance: Our paper Breaking the Curse of Dimensionality: Diffusion Models Efficiently Learn Low-Dimensional Distributions has been accepted at Journal of Machine Learning Research (JMLR).

Jun 2026 Paper Release: 5 new preprints from our group, spanning in-context learning, the representation space of diffusion models, multi-agent communication, long-horizon robotic manipulation, and few-step video generation

May 2026 Paper Acceptance: Our paper Decoupled Data Consistency with Diffusion Purification for Image Restoration has been accepted at ICCP’26.

May 2026 Paper Release: ForcingDAS: Unified and Robust Data Assimilation via Diffusion Forcing

May 2026 Paper Acceptance: Our paper Evaluating the Representation Space of Diffusion Models via Self-Supervised Principles has been accepted at ICML’26.

Apr 2026 Paper Release: Subspace Control: Turning Constrained Model Steering into Controllable Spectral Optimization

Mar 2026 Resource Release: Curated reading list on Memorization and Generalization of Diffusion Models

Mar 2026 Paper Acceptance: Our paper SpaceTools: Tool-Augmented Spatial Reasoning via Double Interactive RL has been accepted at CVPR’26.

Feb 2026 Paper Acceptance: Our paper An Overview of Low-Rank Structures in the Training and Adaptation of Large Models has been accepted at IEEE Signal Processing Magazine (SPM).

Feb 2026 Paper Release: Emergent Low-Rank Training Dynamics in MLPs with Smooth Activations

Jan 2026 Paper Acceptance: 3 papers at ICLR’26, 2 papers at AISTATS’26, 1 paper at CPAL’26

Generalization of Diffusion Models Arises with a Balanced Representation Space (ICLR’26)
AlphaFlow: Understanding and Improving MeanFlow Models (ICLR’26)
Unlearning Isn’t Invisible: Detecting Unlearning Traces in LLMs from Model Outputs (ICLR’26)
Understanding How Nonlinear Layers Create Linearly Separable Features for Low-Dimensional Data (AISTATS’26)
Out-of-Distribution Generalization of In-Context Learning: A Low-Dimensional Subspace Perspective (AISTATS’26)
Analyzing and Improving Model Collapse in Rectified Flow Models (CPAL’26)

OCT 2025 Paper Acceptance: 6 papers have been accepted to NeurIPS'25

3 Spotlights 3 Posters

Spotlight Papers (top 3.2%)

Poster Papers

Apr 2025 SIAM News Article on understanding generalization of diffusion models (news)

Sep 2026 Student Award: Alec Xu received the 2026-2027 J. Robert Beyster Computational Innovation Graduate Fellowship, supporting his work on the efficiency and generalization of deep learning and generative AI models (news)

May 2026 Best Paper Award: Our paper Out-of-Distribution Generalization of In-Context Learning: A Low-Dimensional Subspace Perspective won the Best Paper Award at the AISTATS’26 Workshop on Causality in the Age of AI Scaling

Apr 2026 Student Award: Muhammad Ashiq received the 2026 NSF Graduate Fellowship (GRFP)

Mar 2026 Student Award: Huijie Zhang received the 2026-2027 Rackham Predoctoral Fellowship

Jan 2026 PI Award: Prof. Qing Qu received the 1938E Award in UM College of Engineering (news), a prestigious CoE award for an assistant professor awarded annually to one faculty member for excellence in teaching, mentoring, and service

Jan 2026 Student Award: Can Yaras received a Rising Star Award at CPAL’26

Jun 2025 PI Award: Prof. Qing Qu received the Google Research Scholar Award in machine learning

Mar 2025 Student Awards: Congrats to Can Yaras and Siyi Chen on receiving the Rackham Predoctoral Fellowship! (news1, news2)

Aug 2026 Group Member News: Can Yaras graduated and has joined Princeton Language and Intelligence (PLI) as a postdoc, working with Profs. Sanjeev Arora and Tri Dao. Congratulations, Dr. Yaras!

Jan 2026 Group Member News: Soo Min Kwon successfully defended his thesis. Congratulations, Dr. Kwon! He graduated in May 2026 and has joined Microsoft.

Jan 2026 Group Member News: Xiao Li successfully defended his thesis. Congratulations, Dr. Li! He has joined the University of Hong Kong as a postdoc, working with Prof. Yi Ma.

May 2025 Group Member News: our lab members (1) Yifu Lu will be joining Princeton ECE as a PhD student, (2) Jinfan Zhou will be joining UChicago CS as a PhD student, (3) Pengyu Li will be joining Duke ECE as a PhD student, (4) Siyi Chen will be joining UMich ECE as a PhD student, (5) Dr. Peng Wang will be joining University of Macau as an assistant professor in Fall 2025.

About the Lab

Recent News

Spotlight Papers (top 3.2%)

Poster Papers

Acknowledgement

Visitor Map