3 ms·Skeleton-of-Thought: Prompting LLMs for Efficient Parallel Generation1 points by wseqyrku 2y ago