Hacker News new | ask | show | jobs
by servo_sausage 44 days ago
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.