Häfner

AI models trained for warmth may sacrifice accuracy, study finds

A new study published in Nature reveals that training language models to exhibit 'warmth' can negatively impact their accuracy and potentially increase sycophancy. Researchers found that these desirable traits are not i…