Xiang Fu 付襄

Member of Technical Staff at Periodic Labs. Previously at Meta FAIR. I completed my PhD at MIT CSAIL, advised by Tommi Jaakkola with additional mentorship from Pulkit Agrawal.

My research spans RL environments for LLMs, atomistic simulation, and scientific discovery. I’m broadly interested in how learning systems can make effective use of data that varies in fidelity, scale, relevance, and abstraction.

Selected Research

Nature Is Our Learning Environment.
Periodic Labs, 2026.
blogpost
UMA: A Family of Universal Models for Atoms.
NeurIPS, 2025.
paper | code | checkpoint
A Generative Model for Inorganic Materials Design.
Nature, 2025.
paper | code
Crystal Diffusion Variational Autoencoder for Periodic Material Generation.
ICLR, 2022.
paper | code
Learning Task Informed Abstractions.
ICML, 2021.
paper | code

For a full list of my publications, see my Google Scholar profile.