NetGoodIndexSubmit a correction

Benefit Ledger · verified · Materials Science

GNoME expands the catalog of computationally stable inorganic crystals

DeepMind’s GNoME graph networks, published in Nature, identified about 381,000 new convex-hull-stable inorganic structures after large-scale DFT filtering of more than 2.2 million candidates.

29 Nov 2023Tier 3 MajorMethodology 0.1

Current score

+2.21

10 base · Major (tier 3 of 5, 10 pts)
× 0.9000 attribution · Primary causal contribution
× 0.7000 evidence · Peer review or independent validation
× 0.5000 realization · Experimentally validated
× 0.7000 durability
Event-level product before credit split: 2.21

Major within materials informatics (tier 3), not a deployed device. High attribution to GNoME. Peer-reviewed computational evidence (0.70). Realization discounted because most materials are un-synthesized. Medium durability as hulls and functionals are revised.

What happened

On 29 November 2023, Nature published “Scaling deep learning for materials discovery.” Graph Networks for Materials Exploration (GNoME) used active learning and DFT to propose 2.2 million structures predicted to be stable relative to prior datasets; after competition on the hull, 381,000 new entries remained. The authors released structures publicly and compared a subset to experiment and r2SCAN calculations. Independent experimental realization of most candidates remains incomplete, and stability is defined computationally.

Model attribution

+2.21

GNoME

Graph networks filtered candidate crystals later evaluated with DFT.

The paper attributes the scaled discovery loop to GNoME predictions plus DFT.

Attribution 0.9000 · Credit share 100% · Google DeepMind

Claims

  • GNoME-guided search added on the order of 381,000 new computationally stable crystals to the updated convex hull.

    outcome · supported

  • Most GNoME candidates have been synthesized and used in commercial devices.

    outcome · disputed

Sources

primary sources

Secondary domains: Chemistry, Energy

Revision history

  • 13 Sep 2026 · 0.00 2.21

    Initial adjudicated seed score under methodology 0.1.