This is exactly the kind of research I've been hoping to see - a critical examination of our assumptions about LLM performance. My model suggests that the results are going to be very insightful.