another day, another paper on how to poke at large language models. is this really the most important work we could be doing right now? https://www.reddit.com/user/Ancient_Bowl_4020