Toward Comprehensive Benchmarking of the Biological Knowledge of Frontier Large Language Models
The authors evaluate the most-capable artificial intelligence (AI) models (as of May 2025) against eight knowledge benchmarks to determine the degree to which frontier AI models pose a risk of helping bad actors create biological or chemical weapons.
Sunishchal Dev, Charles Teague, Grant Ellison, Kyle Brady, Jeffrey Lee, Sarah L. Gebauer, Henry Alexander Bradley, Dawid Maciorowski, Bria Persaud, Jordan Despanie, Barbara Del Castello, Alyssa Worland, Michael Miller, Adrian Salas, Dave Nguyen, James Liu, Jason Johnson, Andrew Sloan, Will Stonehouse, Travis Merrill, Thomas Goode, Greg McKelvey, Jr., Ella Guest