1Training large language models on narrow tasks can lead to broad misalignment (opens in new tab)(nature.com)3petemetefete5mo ago1Save
