A model fine-tuned on instruction-response pairs so it follows user prompts more reliably.
Adhering to complex, structured, or constrained instructions