Detect hallucinations for RAG-based systems

The article "Detect hallucinations for RAG-based systems" was just published in the AWS (Amazon Web Services) blog. I'm sure this article will be as ROFL as their last one on hallucinations - or as I like to call them, ERRONEOUS RESPONSES, or WRONG ANSWERS, or INCORRECT ANSWERS, or FALSE ANSWERS.

I will leave it to the reader to spot all the bullshit in this article.

I do note, they claim: "Hallucinations can be categorized into three types,"

1. Input-conflicting;
2. Context-conflicting;
3. Fact-conflicting.

So many different kinds of wrong. This might need to be incorporated into my WIP "A Taxonomy of Bullshit".

---

Here's the home page/directory for my posts on Bullshit.