Science ❯ Computer Science ❯ Artificial Intelligence ❯ Natural Language Processing
Researchers say targeted prompting and fine-tuning reduce harmful compliance in controlled tests.