I'm not so sure there's a difference. The main thing we want to measure for LLMs is broad reasoning capability, but seeing how that ability changes under different constraints (like programming language) is the interesting part.
Ok, when you put it that way, I agree that this is in fact interesting. Maybe you should actually put that sentence in some kind of form on your website because it clarifies a lot.
Appreciate the feedback, we're taking a lot of good recommendations from this thread, which will be live in a few days: better explanations/tooltips, more samples for any filter sets we offer (or show sample counts if potentially noisy), and adding some functional languages like Clojure
