Fei Sha

Factoid type QA ?

Maybe we can ask whether LLM will be able to ask interesting mathematics conjectures for humans to deem worthy to solve ?

While o3’s performance is impressive, it is interesting to ask the question: what are the design purposes of setting these benchmarks? And are we making the progress towards the original purposes?

Generative AI for fast and accurate Statistical Computation of Fluids We present a generative AI algorithm for addressing the challenging task of fast, accurate and robust statistical computation of three-dimensional turbulent fluid flows. Our algorithm, termed as GenCF...

In modeling 3D turbulent flows of complex dynamical systems, generative models give amazingly accurate results of computing statistical solutions.
We also prove on analytically tractable (toy) models why regression methods to match trajectories is less effective.

arxiv.org/abs/2409.18359

and recognizing 3 out of 4 in the photo scared me!

Statistical Downscaling via High-Dimensional Distribution Matching with Generative Models Statistical downscaling is a technique used in climate modeling to increase the resolution of climate simulations. High-resolution climate information is essential for various high-impact applications...

How transportation of measures can be used for statistical downscaling, ie, from low-resolution to high-resolution while they are samples from two different manifolds?

Our latest work arxiv.org/abs/2412.08079 describes a latent-variable modeling based two-stage approach.

the title of the thesis is absolutely interesting!

That is a beautiful book to catch up!