Hi all,
I’m here to talk about the coherence of ChatGPT’s output with thread information. I’m using ChatGPT 4, and when threads become longer, it seems like the model can’t find previous data and outputs, causing it to start hallucinating. For example, if ChatGPT creates a story with a timeline and the conversation continues, after a while, it starts forgetting the original story, dates, and characters. Similarly, if I’m creating a to-do list and we continue adding details, after some prompts, ChatGPT completely forgets the initial list and starts creating new items that are different from the original. When you try to make it “remember” or reread the thread, it fails. Do you know how to avoid this? Are there any best practices to help ChatGPT remain coherent?
thank you