OpenAIStory of the weekResearch paper

Improving language model behavior by training on a curated dataset

OpenAI research finds that fine-tuning on a small, curated dataset can improve language model behavior with respect to specific behavioral values.