Hacker News new | ask | show | jobs
by SomaticPirate 196 days ago
Are these benchmarks correct that adding Anthropic's Constitutional AI system prompt lowered results across all the models?