I love the ability to rewrite a previous message and change the context from that point forward, but what would really help me is if I could simply remove parts of the prior conversation. There are times where hallucinations aren’t caught until later, and I’d like to be able to go back a few messages and remove a part of the context without losing everything that came after it. I’ve tried chatting things like “well that part from earlier was wrong but the rest was correct” but it never works out. It doesn’t seem like the LLM can understand partial issues, so being able to remove the incorrect part would be fantastic.
3 Likes
Great suggestion. I had similar experiences where the AI goes wrong after making one small assumption. Removing that would help it be back on track.
Sending a correction prompt to tell AI that the assumption or change was wrong doesnt prevent it from further hallucinating about that part. specially if context is getting longer.