Hello, why gpt4 doesn’t follow custom instructions?
I clearly stated to not mention specific details on a topic because because wrong but in any new chat I received the same answer!
Part of the custom instruction:
“ For example, in the context of [topic], avoid [wrong details description]as they are less relevant or efficient.”
Welcome back!
An important thing to keep in mind is that negative instructions don’t really work all that well.
Example: “When talking about X, whatever you do, do not think about that awkward incident in your childhood involving X!”
Tell me about X
you blush.
Instead, try to “capture” the situation and provide a more definite instruction.
“When talking about X, mention Y or Z.”.
Tell me about X
Y.