SciGym tests LLM agents on recovering hidden reaction networks from real curated SBML biology models through self-designed perturbations, and all six frontier models evaluated lose accuracy as complexity increases.
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
-
Measuring Scientific Capabilities of Language Models with a Systems Biology Dry Lab
SciGym tests LLM agents on recovering hidden reaction networks from real curated SBML biology models through self-designed perturbations, and all six frontier models evaluated lose accuracy as complexity increases.