Necessary conditions for benchmarking next-generation AI in materials discovery

Daniel E Widdowson, Tatiana V Kurlina, Olga D Anosova, Vitaliy A Kurlin

arXivarxiv/2610.13885Oct 9, 2026cond-mat.mtrl-sciscore 0

Current AI predicts materials by speeding up energy minimisation and training on large databases. However, scaling data generation prevents manual analysis by domain experts. If novelty is evaluated by chemical compositions or traditional representations in Crystallographic Information Files, or standard structure matchers based on manually chosen thresholds, one can easily generate millions of ``new'' compositions or files without discovering truly novel materials. This paper uses a hierarchy of crystal invariants to continuously trace and quantify structural novelty based on the new…

SourceJSON

Comments and reviews

No comments yet.

API

curl /api/v1/records/arxiv/2610.13885
curl -X POST /api/v1/records/arxiv/2610.13885/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2610.13885/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"body": "…", "review": true}'