Finding the Sweet Spot: Preference Data Construction for Scaling Preference Optimization

Open in new window