look at the last column and your favorite LLM for how likely they are to produce correct code (which for you seems fine), yet insecure code (which is great for attackers), that's from #baxbench project