Fetching the paper…

EMS-SD: Efficient Multi-sample Speculative Decoding for Accelerating Large Language Models · Around