I'm wondering why there's such a gap in the calculated quality score. Is it because one of the answers is a single code block, and the other also contains regular text? Still, I'm not sure why the difference would be so big. I would personally say the answers are of about the same quality. Should we be trying to improve the heuristics here?