Suppose we want to use ChatGPT to create an assistant that only responds with information contained in a SOURCE document.
How can we implement a fact-checking method? Specifically, how can we instruct ChatGPT to verify that the response contains only information from the SOURCE file? (using commands included in the prompt)
Only way I can think of at the moment with ChatGPT would be to use an Action to run the output through an LLM on your own server, but even then it wouldn’t be 100% likely.