BARE: Combining Base and Instruction-Tuned Language Models for Better Synthetic Data Generation

Open in new window