No, it does not. You said above that "The LLM gives you the answer it finds on the training set" You and I both agree on that. No first principles there. The

No, it does not. You said above that "The LLM gives you the answer it finds on the training set" You and I both agree on that. No first principles there. The training set -- how's it built? With tokens. We have not trained LLMs with a token structure that deals well with compound words. If we trained LLMs with a different token structure, it is more probable that a one-shot answer for these compound word letter counting problems would be accurate. The LLM does not need to understand what "counting is" or even "what a letter is". The LLM will regurgitate the token relationship we train it on.