PapersSelected72°07:42Chinese contentTokTier有状态分词提升智能体服务,TTFT最高降34%做智能体服务的朋友看下,TokTier把长记录重复分词干掉了,vLLM下首字延迟降16%-34%,省GPU。#TokTier#vLLM#智能体#分词elvis@omarsar0SourceRead laterReadWorth followingUsefulFollow TokTier