Llama 2
Meta
rich detail 21 method fields published
Evaluation
Safety human evaluation (adversarial prompt battery)
Human-rated safety violations on roughly 2,000 adversarial prompts (1,351 single-turn, 623 multi-turn) across safety categories; 5-point Likert safety scale; per-category violation percentages
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.