[ICLR 2024] Skeleton-of-Thought: Large Language Models Can Do Parallel Decoding - View it on GitHub