Prompt-Robust Language Models: Which Training Strategies Work?

Frederic Sadrieh, Michal Štefánik

Despite their strong performance, large language models remain highly sensitive to prompt formulation. Prior work addresses this through refined data construction or through dedica…