[dotancohen]

Where does fine-tuning sit in this? How easily are existing models able to be fine-tuned for new use cases, such as specifically legal or medical texts?