Next-Token Prediction Task Assumes Optimal Data Ordering for LLM Training in Proof GenerationChenyang An, Maurice Diesendruck, Shima Imani et al.|arXiv (Cornell University)|2024Cited by 0