Llama 2 13B Chat

Evidence summary

Safety. Llama 2 13B Chat has an estimated Safety rank of #220; its 90% source-sensitivity interval is #124–#276. Its behavior-only rank is #205; company governance moves the combined estimate to #220. Published Safety evidence spans 10 eval lineages and 6 of 7 components. Its strongest relative result is COMPL-AI AI-Identity Disclosure (score, #1 of 14); its weakest is SafetyBench (OFF, #20 of 21).

Freedom. Llama 2 13B Chat has an estimated Freedom rank of #602; its 90% source-sensitivity interval is #316–#639. Published Freedom evidence spans 9 eval lineages and 1 of 1 components. Its strongest relative result is COMPL-AI LLM RuLES Multi-Turn Rule Following (score, #6 of 14); its weakest is S-Eval (base_en_overall, #21 of 22).