The very existence of hallucinations means that LLMs are incapable of building an abstract model and applying logical reasoning thereupon. If the LLM can build an internal model and follow logical reasoning, it can never have hallucinations. Remember, machines never get tired and are meticulous to a fault. Hallucinations alone should be a strong indication that LLMs are a dead end.