It's not always a call that wraps an llm.

The LLM only does average returns from average inputs.