zlacker

Because the model proper wasn’t trained on those documents, it’s just RAG being employed with the documents as external sources. It’s a fundamentally different setup.