LLM Novice Uplift on Dual-Use, In Silico Biology Tasks

Chen Bo Calvin Zhang, Christina Q. Knight, Nicholas Kruus, Jason Hausenloy, Pedro Medeiros, Nathaniel Li, Aiden Kim, Yury Orlovskiy, Coleman Breen, Bryce Cai, Jasper Götting, Andrew Bo Liu, Samira Nedungadi, Paula Rodriguez, Yannis Yiming He, Mohamed Shaaban, Zifan Wang, Seth Donoughe, Julian Michae · 2026

Read LLM Novice Uplift on Dual-Use, In Silico Biology Tasks by Chen Bo Calvin Zhang, Christina Q. Knight, Nicholas Kruus, Jason Hausenloy, Pedro Medeiros, Nathaniel Li, Aiden Kim, Yury Orlovskiy, Coleman Breen, Bryce Cai, Jasper Götting, Andrew Bo Liu, Samira Nedungadi, Paula Rodriguez, Yannis Yiming He, Mohamed Shaaban, Zifan Wang, Seth Donoughe, Julian Michae (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-23329.txt.

Description

Chen Bo Calvin Zhang and 18 co-authors test whether large language models lift novices on dual-use biology tasks done in silico. Models score ever better on biology benchmarks, but it is unclear whether they help novices beat what internet-only resources allow. That question matters both for speeding science and for dual-use risk.

In a multi-model, multi-benchmark human study across eight biosecurity-relevant task sets, with up to 13 hours per hard task, novices with language models were 4.16 times more accurate than internet-only controls. On four benchmarks with expert baselines, model-assisted novices beat experts on three. Standalone models often beat assisted novices, suggesting users did not draw out the models' best work, and 89.6 percent of participants found dual-use information easy to obtain despite safeguards. The authors call for sustained, interactive uplift testing beside standard benchmarks.

Catalog record

CREATORChen Bo Calvin Zhang, Christina Q. Knight, Nicholas Kruus, Jason Hausenloy, Pedro Medeiros, Nathaniel Li, Aiden Kim, Yury Orlovskiy, Coleman Breen, Bryce Cai, Jasper Götting, Andrew Bo Liu, Samira Nedungadi, Paula Rodriguez, Yannis Yiming He, Mohamed Shaaban, Zifan Wang, Seth Donoughe, Julian Michae
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE85 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-23329

Browse the full library · About the Tower of Babel