Frontier LLMs Match Human Experts on Biological Phenotype Annotation, Offering Path to Scale Scientific Data Curation

Researchers tested five advanced language models from Anthropic and OpenAI as autonomous curators for annotating biological phenotypes—a traditionally labor-intensive task requiring expert humans. All five agents performed within the range of expert human curators on a standardized benchmark and significantly outperformed prior NLP tools, suggesting LLM-based agents could help overcome a major bottleneck in comparative biology research.

Why it matters: As life sciences increasingly rely on integrated cross-study data, automating phenotype annotation at human-expert level could dramatically accelerate research while reducing dependence on scarce trained curators.