Google DeepMind trained Gemini 1.5 Flash to adopt specific values by midtraining it on synthetic documents before chat finetuning. This two-step process ensures traits remain robust even in out-of-distribution scenarios. The team's findings provide a concrete blueprint for aligning model behavior. Practitioners can now better control specific personality traits through targeted synthetic data generation.