A new LLM benchmark of 1,757 graduate-level materials science questions shows proprietary models outperform open-source and domain-specific models, though the benchmark's LLM-generated gold answers raise validity concerns.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
MSQA: Benchmarking LLMs on Graduate-Level Materials Science Reasoning and Knowledge
A new LLM benchmark of 1,757 graduate-level materials science questions shows proprietary models outperform open-source and domain-specific models, though the benchmark's LLM-generated gold answers raise validity concerns.