Llama 2
Meta
partial detail 10 method fields published
Evaluation
Ghost Attention (GAtt) multi-turn consistency evaluation
Quantitative analysis of GAtt (applied after RLHF V3) consistency up to 20+ turns until maximum context length, including inference-time constraints like 'Always answer with Haiku'
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.