Skip to content

Evaluation

Context distillation with answer templates (safety)

Impact of context distillation and context distillation with risk-category answer templates on safety reward model scores

Model behavior & alignment Meta

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.

Llama 2
Meta

partial detail 12 method fields published

See how this test was run