OpenAI：小规模精选数据微调可改善模型行为

Our latest research finds we can improve language model behavior with respect to specific behavioral values by fine-tuning on a small, curated dataset.