Distillation
Beginner
Teaching a small AI by letting it copy a big AI’s answers.
Novice
Training a smaller student model on outputs generated by a larger teacher model, so the student picks up the teacher’s skills at a lower running cost.
Expert
Here, supervised fine-tuning of a student on teacher-generated reasoning traces (sequence-level distillation), not matching the teacher’s logits. It transfers behavior the student is unlikely to discover with RL on its own.