Yeah wouldn’t an llm just assume it was a typo and just match it with the closest word it thinks it is like it would do when you make a minor typo in a prompt anyway?