Llama 2
Meta
partial detail 12 method fields published
Evaluation
Safety data scaling ablation (RLHF data mix)
Ablation adjusting the amount of safety data in the RLHF stage (helpfulness data fixed); mean reward model scores proxy safety and helpfulness performance
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.