The Multilingual Alignment Prism: Aligning Global and Local Preferences to Reduce Harm

Open in new window