For the past 72 hours, using 4o, 4.5 or Legacy 4 in ChatGPT (Web, Mobile, App) is resulting in consistent hallucinations for details given in project files.
It’s always been 90-95% accurate when it comes to file usage but now it’s dropped to a 50%. Where I have to double check what it is saying in my own files—which is, far from ideal.
Yes. Having the same issue. I have several Custom GPT’s with uploaded knowledge files which they no longer seem to be using. Instead they hallucinate in every reply.
I have the same issue as well, I have even copied all the text from pdfs, and created a .txt file and I still get hallucinations. I am using it to help me study for my finals coming up and and it is for adolescent development and it has only given me unrelated response. Some are notes on how a joint works and its parts, the muscular skeletal system, supply demand the, origins of progressivism, the industrial revolution and other completely unrelated topics I have tried everything. Nothing works.