BioAgent Bench is a new evaluation suite that tests AI agents on end-to-end bioinformatics pipelines and finds that frontier models often complete tasks reliably but fail under controlled perturbations like corrupted inputs or prompt bloat.
Inputs (provided at evaluation time) •1
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
BioAgent Bench: An AI Agent Evaluation Suite for Bioinformatics
BioAgent Bench is a new evaluation suite that tests AI agents on end-to-end bioinformatics pipelines and finds that frontier models often complete tasks reliably but fail under controlled perturbations like corrupted inputs or prompt bloat.