Comment by jameslk

9 hours ago

Could models mitigate this by answering questions incorrectly with random information instead of outright refusing to answer them?

0 comments