The mistakes they make are very similar to mistakes we humans do. Just like you can confuse a human with irrelevant information, you can distract the LLM. We ar

The mistakes they make are very similar to mistakes we humans do. Just like you can confuse a human with irrelevant information, you can distract the LLM. We are not good at big tables of integers, just as them. An LLM isn't a calculator. But we probably can teach it how to use one.