AI models can judge if their own answers are correct

Original: Language Models Mostly Know What They Know

Source: anthropic.com ↗