This is very cool. It seems like the prompt is asking the LLM to one shot an answer. Have you tried asking it to make a group, confirm whether it's correct, and

This is very cool. It seems like the prompt is asking the LLM to one shot an answer. Have you tried asking it to make a group, confirm whether it's correct, and repeat with the remaining words? (like a human would)