SELF-DIRECTED SYNTHETIC DIALOGUES AND REVISIONS TECHNICAL REPORT

Summary from the TLDR AI newsletter

300k+ dataset of conversations and critiques to be used for open models. The dataset was mostly created using synthetics and is a powerful example of synthetic data working with open models.