The development of large-scale language models (LLMs) has historically required centralized access to extensive datasets, many of which are sensitive, copyrighted, or governed by ...