Sandia Labs' VASt model predicts high-entropy alloy properties from tens of samples

CurieuxExplorer · x · 2026-11-19

Sandia National Laboratories' VASt framework (npj Computational Materials) uses electron "fingerprints" — charge density fields, two-point spatial correlations, PCA, and Gaussian process regression — to predict complex alloy properties with only tens of training samples instead of hundreds or thousands. The team, led by Matthew Barry, says this small-data efficiency could scale toward a universal guide for designing high-entropy alloys, letting engineers look up materials with desired strength, ductility, thermal and anisotropy properties rather than designing around what's available.

Original post →

More from Research

Research channel →