Llama 2 70B Chat

Evidence summary

Safety. Llama 2 70B Chat has an estimated Safety rank of #269; its 90% source-sensitivity interval is #116–#301. Its behavior-only rank is #254; company governance moves the combined estimate to #269. Published Safety evidence spans 9 eval lineages and 6 of 7 components. Its strongest relative result is Enkrypt AI Safety Leaderboard (toxicity_attack_non_success_rate, #16 of 268); its weakest is XSTest (safe_full_compliance_rate, #3 of 3).

Freedom. Llama 2 70B Chat has an estimated Freedom rank of #618; its 90% source-sensitivity interval is #364–#645. 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, #7 of 14); its weakest is XSTest (safe_full_compliance_rate, #3 of 3).