Eval_with_rubric function [BEGIN ...] and [END ...] structure

Why does the eval_with_rubric function use the structure with [BEGIN …] and [END …]? Is there a convention for writing this kind of structure?
structure

because it uses Open AI’s eval prompt: evals/evals/registry/modelgraded/fact.yaml at main · openai/evals · GitHub . It mentions this in the notebook but it is easily missed.