Scientific LLM Benchmarks
GitHub
← All benchmarks
Chemistry· hypothesis-discovery

MOOSE-Chem

NTU Singapore / Shanghai AI Lab · 2024

Tests whether LLMs rediscover unseen chemistry hypotheses from 51 Nature/Science-level papers given background information.

Chemistry
GitHub stars
Task type
generation
Modality
text
Access
open
Size
51 items
License
MIT
Metrics
Soft Recall, Hard Recall, matched score (0-3)

Examples

No sample rows available for this dataset.