Yes, and corrections stick. Reject a bad draft or thumbs-down an answer and Heidi corrects the memory behind it, with the reason on record. A correction teaches the brain, not just the chat.
In most AI tools, telling the model it is wrong fixes one conversation. In Heidi, a correction retires or fixes the underlying fact, so the whole company stops getting the wrong answer, permanently.
Every correction is auditable: what changed, when, and why.