DSpark: Speculative decoding accelerates LLM inference [pdf]

Front page story about speculative decoding for LLM inference.