SafeDPO: A Simple Approach to Direct Preference Optimization with Enhanced Safety

Open in new window