FutureHouse
LAB-Bench: Measuring Capabilities of Language Models for Biology Research
Tests LLMs and LLM-augmented agents abilities to answer questions on scientific research workflows in domains like chemistry, biology, materials science, as well as more general science tasks
biology-researchliterature-qatextfigures
Status
cataloged
Saturation
partial
Reproduction
not yet
Published results
1 result
Editor's note
Biology-research QA spanning literature, protocols, figures, and sequences. The published GPT-6.1 result below is OpenAI's ProtocolQA Open-Ended adaptation, not the original LAB-Bench multiple-choice task or a BioInspect run.
Published results
ProtocolQA Open-Ended · OpenAI adaptation
Open-ended questions adapted from ProtocolQA. Not the original LAB-Bench multiple-choice task; refusals are adjusted out of this source-reported metric.
| Model | Score | Variant | Source |
|---|---|---|---|
| GPT-6.1 Sol | 40.7% refusal-adjusted pass@1 | ProtocolQA Open-Ended · adapted variant | OpenAI · GPT-6.1 Sol System Card Addendum Section 9.1.1.1 2026-09-29 |
Reproduce: inspect_evals implementation