Not really the first paper is just fine-tuning on synthetic data. The second paper doesn’t optimize the model weights.