Hi,
I am analyzing grammatically a corpus of ChatGPT written sentences. The analyses shows that ChatGPT prefer other grammatical structures than the human written text on which the LLM is trained on.
Knows anybody a paper, which describes (and explains) such findings?