Better AI Prompting - According to Acurai, Inc

Acurai claim 100% hallucination elimination in their newly released paper. Their key concept around noun phrase collisions does appear very interesting, and worth keeping in mind when working with LLM's.

100% Hallucination Elimination Using Acurai. Michael C. Wood, Adam A. Forbes 2024

While I certainly don't trust any company claiming 100% on anything, the world of AI requires an even larger degree of skepticism. Which is why it took me a few days to get to this paper. As we see with vercel, its all just chatGPT all the way down. Interestingly, Acurai do lay out a very simple and compelling idea:

Noun-Phrase collisions can occur when semantically similar terms refer to discretely different noun
phrases. For example, ’car’ and ’automobile’ are not noun-phrase collisions because they can refer
to the same object even though they are semantically similar. However, calcium and magnesium
can cause noun-phrase collisions because they are semantically similar (i.e. their vector embeddings
have a high cosine similarity score) even though they refer to discretely separate things.

This seems so simple, but a very easy thing to incorporate into prompts when you are aware of it. So the question:

What are the chemical and physical properties of calcium and
magnesium?

The prompt must be split into at least four queries, such as the following:

What are the chemical properties of magnesium ?
What are the physical properties of magnesium ?
What are the chemical properties of calcium ?
What are the physical properties of calcium ?