Improving language model behavior by training on a curated dataset

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

Source: OpenAI — Published June 10, 2021 at 07:00 UTC — Category: Models