HADES: Hardware Accelerated Decoding for Efficient Speculation in Large Language Models

Ze Yang, Yihong Jin, Xinhe Xu

Large Language Models (LLMs) have revolutionized natural language processing by understanding and generating human-like text. However, the increasing demand for more sophisticated…