Llama 2
Meta
partial detail 11 method fields published
Evaluation
Red teaming insight analysis and robustness measurement
Analysis of red teaming data (dialogue length, risk area distribution, misinformation topics, rated risk) and robustness gamma = average prompts triggering a violating response per person per hour; percentage of previously discovered violating prompts rejected by new models
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.