How are complete nonsense papers being accepted to top conferences? [D]

I saw someone's paper get accepted to ICML and they literally just did prompt engineering. Ignoring whatever misgivings about how pointless that is on its own, they didn't even have any inkling of a reasonable design. They evaluated on 10 samples, used another llm to evaluate, used an arbitrary heuristic to decide said prompts, and had a terribly unrealistic testing scenario. I'm not an LLM person and I see papers in my field not only with amazing results, experimental design, and strong theore