Hao AI Lab· @haoailab · X·· 2 小时前AI 评分31
AI 导读
[3/5] 等待验证结果的剩余空闲时间，现在全部用于打磨更好的草稿。在目标模型验证期间，JetSpec++ 会修正自己的并行草稿：它依据自身猜测重新起草，刷新预测的上下文，并在由目标模型最后几层构建的解码器出现分歧的位置修正第一个 token。 在 7 个基准上，每轮平均接受 8.94 个 token，而 DSpark 为 6.89 个。
正文 · 原文
[3/5] The rest of spare time waiting for verification outcome now goes into refining a better draft. While the target verifies, JetSpec++ corrects its own parallel draft: it redrafts on its own guesses, refreshes its predicted context, and fixes the first token where a decoder built from the target's last layers disagrees.
8.94 accepted tokens per round across 7 benchmarks, vs 6.89 for DSpark.
来源：Hao AI Lab · x.com