GPT4o worse that 4 in reading files

OK, so the thing goes like this: I feed my GPT with a file with a list of exam questions. It is supposed to just go one by one and not give me an answer until I answer. It’s ok with the latter. But it’s just stopped to be able to give correct questions. I deleted it several time. I have even feed it with a smaller base right now - and it turns out he can’t even start with the right question right now. It just invents questions instead of reading and pasting them.

What’s more - it wasn’t even able to understand that it gives me the same question all over even though I’m telling it that this is the wrong question. When I asked itif the last one is different that the previous it had given me, it answered that they differ, cause they are the same…


I understand your frustration. I’m experiencing the same. Quality of output is horrendous. I just want to know if the developers will fix it or it’ll continue degrading in quality. And, if there is a temporary prompting solution to make it act “intelligently” like it used to before.