x
Training goals for large language models — AI Alignment Forum