cd/entity/CARAT· home› entities› CARAT
grep -l @carat /news/*.json | wc -l → 1

CARAT

mentions 1 type Organization feed RSS

// recent coverage 1 mentions

04:00
2026-10-01
arxiv.org
ai-research

CARAT: Do Materials LLMs Reason or Recite?

The CARAT benchmark, detailed in arXiv paper 2609.38340v1, finds that materials LLMs often recite structural relations rather than reason from them: a frozen model quoted a link yet answered identical…

// co-occurs with top 2 entities
GraphSpace 1 arXiv 1
// topics top 4 topics
ai research 1 large language models 1 machine learning 1 artificial intelligence 1