The Mistral 7b Self-Alignment SFT Model is an adapter fine-tuned specifically for self-alignment and harmlessness. It has been trained using the Mistral Self-Alignment Preference Dataset, which can be accessed here.

The fine-tuning process is detailed on the corresponding GitHub page, providing insights into the methodology and purpose behind the model's adaptation.