A arxiv.org Simple tasks showing reasoning breakdown in state-of-the-art LLMs 375 points tosh 2 years ago 380 comments
