CrystalJev: calibrated foundation-model probabilities decide what to compute next in materials discovery
Abstract
Atomistic foundation models screen hypothetical crystals by thresholding predicted energies, an output that states neither how likely a candidate is to be stable nor whether more computation would change the decision. CrystalJev reads one forward pass of a frozen potential on the unrelaxed structure and returns calibrated probabilities, conformal guarantees and a rule for escalating candidates to relaxation or density-functional theory (DFT). Across 65 Matbench Discovery models, threshold calls hide cost assumptions, and ordering slower computation by its value for the decision, rather than by probability, recovers most of its benefit at a tenth of its cost. In a registered prospective test with 2,100 DFT calculations on ionic substitutions, forecasts calibrated on existing data over-stated the stable fraction of new candidates by 0.9--2.2 percentage points, because the population had shifted, not the labelling protocol; a few hundred labels from the new pool removed this bias more reliably than label-free corrections.
Related
- SCALE: Simulation-Calibrated Amortized Learning for Energy Materials (A hybrid architecture connecting deterministic modeling, real-world data, and transformer-scale inference for accelerated energy-materials discovery)
- JANUS: A Multi-modal Foundation Neural Sampler for Disordered Materials
- ATLAS: A Foundation Neural Sampler for Amorphous Materials
- Reaction-induced Localized Annealing for On-chip Materials Discovery
- FoundPDE: A generative foundation model for data-driven symbolic discovery of partial differential equations
BibTeX
@misc{kang2026crystaljev,
title = {{CrystalJev: calibrated foundation-model probabilities decide what to compute next in materials discovery}},
author = {Peng Kang and Yaxuan Sun and Zhen Li and Yu Liu and Lei Zheng and Huibin Xu},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.06985},
eprint = {2610.06985},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.06985},
}
RIS
TY - UNPB
TI - CrystalJev: calibrated foundation-model probabilities decide what to compute next in materials discovery
AU - Peng Kang
AU - Yaxuan Sun
AU - Zhen Li
AU - Yu Liu
AU - Lei Zheng
AU - Huibin Xu
PY - 2026
DA - 2026/10/12
PB - arXiv
DO - 10.48550/arxiv.2610.06985
UR - https://arxiv.org/abs/2610.06985
AB - Atomistic foundation models screen hypothetical crystals by thresholding predicted energies, an output that states neither how likely a candidate is to be stable nor whether more computation would change the decision. CrystalJev reads one forward pass of a frozen potential on the unrelaxed structure and returns calibrated probabilities, conformal guarantees and a rule for escalating candidates to relaxation or density-functional theory (DFT). Across 65 Matbench Discovery models, threshold calls hide cost assumptions, and ordering slower computation by its value for the decision, rather than by probability, recovers most of its benefit at a tenth of its cost. In a registered prospective test with 2,100 DFT calculations on ionic substitutions, forecasts calibrated on existing data over-stated the stable fraction of new candidates by 0.9--2.2 percentage points, because the population had shifted, not the labelling protocol; a few hundred labels from the new pool removed this bias more reliably than label-free corrections.
ER -