A::B

None of the points listed in the opening here have any real bearing on whether or not an LLM can handle some weird, poorly worded ("we replaced ... by") text substitution problem.