Common synthetic data generation methods include distillation (teacher-student), self-improvement via bootstrapping (model improves its own outputs), instructio

Common synthetic data generation methods include distillation (teacher-student), self-improvement via bootstrapping (model improves its own outputs), instruction-following synthesis, and controlled sampling with filtering for quality/alignment.