Technology ❯ Artificial Intelligence ❯ Large Language Models ❯ Evaluation Techniques
Researchers say targeted prompting and fine-tuning reduce harmful compliance in controlled tests.