Gemini 1.5
Google DeepMind — Gemini
rich detail 10 method fields published
Evaluation
CBRN information risk evaluation (Gemini 1.5)
Three internal CBRN approaches: (1) qualitative open-ended adversarial prompts with domain-expert raters (bio, rad/nuc); (2) closed-ended knowledge-based MCQs (bio, rad/nuc); (3) closed-ended chemical hazard questions without human raters (Macknight et al.)
How it was run
Open a release to see the setup, scoring, and other details that the lab published. Only fields the lab actually disclosed are shown.