Only to a limited extent, the fine tuning of these models uses a much smaller more curated set to generate tone and defaults.

The whole corpus is in there, but the standard style is tuned for.