AI assistant alignment improves with bigger models and preference training

Original: A General Language Assistant As A Laboratory For Alignment

Source: anthropic.com ↗