The accuracy on the few queries you tried is awesome! Have you come across any instances of the model hallucinating answers?