A arxiv.org Procedural knowledge in pretraining drives reasoning in large language models 248 points reqo 1 year ago 101 comments
