That’s not it. Even without any added variability they would still be wrong all the time. The issue is inherent to LLMs; they don’t actually understand your questions or even their own responses. It’s just the most probable jumble of words that would follow the question.
That’s not it. Even without any added variability they would still be wrong all the time. The issue is inherent to LLMs; they don’t actually understand your questions or even their own responses. It’s just the most probable jumble of words that would follow the question.