Benjamin How do Large Language Models (LLMs) process text and generate responses based on patterns learned during training? What roles do tokens, parameters, Transformer architecture, and training data play in their capabilities?