Hacker News new | ask | show | jobs
by trolan 1051 days ago
The model itself was fine tuned for JSON function responses, they admitted that openly. They also acknowledge they make changes to ChatGPT all the time, which has nothing to do with the model underneath it.