Decision models like Jev don't beat LLM-as-a-judge or traditional classifiers