🔂 We ran Self-Play fIne-tuNing (SPIN) on the DIBT prompt collective data so we figured writing a blog with @mantisnlp would be an awesome way to get the community ready for yet another awesome model release that highlights the need for high-quality data.