Instruction Tuning - How does instruction fine-tuning work?

Learning for Multiple Tasks at Once: A Progressive Approach Instruction in multitasking A new paradigm is embraced by fine-tuning, which trains language models on many tasks at once. We guide the model's behavior during fine-tuning by providing explicit instructions for each task, rather than fine-tuning it one task at a time.