In the context of LLMs, which statement BEST characterizes an instruction-tuned Large Language Model?