arxiv preprint - The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits | AI Breakdown | Podwise