News

[2026.09] PAPER Congratulations to Jiacheng Qiu, Yunsoo Kim, and our collaborators on their new preprint! Their student-curriculum coupling approach improves temporal video grounding with fewer training examples. [PDF]

[2026.09] PAPER As first author, I am pleased to share SCOPE, a single-pass approach to recovering full PDE fields from sparse observations. [PDF] [Code]

[2026.09] PAPER I led PDE-OBS as first author; the new preprint benchmarks PDE models across changing observation patterns. [PDF] [Code]

[2026.09] PAPER Congratulations to Haochun Wang and the team on their TMLR acceptance with minor revision! Their time-series forecasting paper studies principled test-time adaptation. [arXiv]

[2026.08] SERVICE Accepted the invitation to serve as a reviewer for ICLR 2027.

[2026.08] PAPER My first-author paper HP-JEPA is now on arXiv. It learns predictive graph representations across multiple resolutions. [PDF]

[2026.06] SERVICE Serving as a reviewer for Transactions on Machine Learning Research (TMLR).

[2026.06–07] GROUP Mentored AI4Science projects in the AI3 REU program at Stony Brook University.

[2026.06] PAPER Congratulations to Georgios Kementzidis and collaborators on their Computer Physics Communications publication! Their multistep generative backmapping method restores atomistic detail from coarse-grained structures.

[2026.05] PAPER Congratulations to Haochun Wang and colleagues on sharing their preprint! Their test-time adaptation study calibrates time-series forecasts without retraining the underlying model. [PDF]

[2026.05] AWARD Recognized as a Gold Reviewer at ICML 2026 for the quality of submitted reviews.

[2026.01] PAPER Congratulations to Jingxiang Qu, Wenhan Gao, and their coauthors on ICLR 2026! Their GAGA paper makes 3D molecular generation more efficient through Gaussianity-aware approximation. [arXiv]

[2026.01] TEACHING Taught AMS 361: Applied Calculus IV at Stony Brook University.

[2025.10] PAPER Congratulations to Wenhan Gao and the research team on their arXiv preprint! They ask whether interpretable neural operators can reveal hidden physical dynamics. [PDF]

[2025.08] PAPER Three papers were accepted to NYSDS 2025 (one oral, two posters). I am first author of the Hamiltonian learning paper and co-corresponding author of the boundary-informed PINN and physics-informed active-learning papers. [arXiv]

[2025.07] PAPER I presented my first-author RL-QESA work as a poster at AI for Math @ ICML 2025, exploring reinforcement-learned cooling policies for simulated annealing.

[2025.06] PAPER Our first-author APOD paper was accepted to the Assessing World Models Workshop @ ICML 2025; it uses observations and PDE constraints to guide diffusion sampling.

[2025.06] PAPER Congratulations to Wenhan Gao, Jian Luo, and colleagues on their TMLR publication! Their Dynamic Schwartz–Fourier Neural Operator uses input-dependent Fourier kernels for greater expressive power.

[2025.05] PAPER I am first author of the new velocity-inferred Hamiltonian networks preprint, which learns energy-conserving dynamics from position-only observations. [PDF]

[2025.05] PAPER As co-first author and project lead, I am glad to share the IJCNN 2025 acceptance of Kolmogorov–Arnold Representation for Symplectic Learning. [arXiv]

[2025.04] PAPER My first-author preprint on move schemes in simulated annealing is available on arXiv, examining how the number of coordinates updated per move affects optimization. [PDF]

[2025.01] PAPER Congratulations to Wenhan Gao and collaborators on their ICLR 2025 paper! Their study of discretization mismatch probes how neural operators behave when the grid changes.

[2024.10] PAPER Congratulations to Wenhan Gao and his coauthors on their TMLR publication! Their Coordinate Transform Fourier Neural Operators incorporate physical symmetries through coordinate transformations.