3 ms·
Could this be to avoid variants of the “do it or I’ll kill you” adversial prompts? Those were pretty effective on OpenAI models.
by ungreased0675 3y ago
Could this be to avoid variants of the “do it or I’ll kill you” adversial prompts? Those were pretty effective on OpenAI models.