It seems the debate seems to center around whether language models are meta-models (in the category sense) or mere encodings (information theory)?
It seems the debate seems to center around whether language models are meta-models (in the category sense) or mere encodings (information theory)?