Polymathic AI 与 Princeton 发布 FlashMap，实现 6.14 倍 LLM 解码加速
Polymathic AI and Princeton Researchers Release FlashMap, a Flow Map Drafter for 6.14x Faster LLM Decoding
Princeton 那边搞了个 FlashMap，一次能猜 16 个 token，Qwen3-8B 解码快了 6.14 倍，推理提速可以看看
Polymathic AI、Princeton 和 USC ISI 的研究者发布了推测解码起草器 FlashMap。FlashMap 采用流映射方法，单次前向传播即可联合起草 16 个 token。在 Qwen3-8B 上，它实现了最高 6.14 倍的无损加速。
Polymathic AI and Princeton Researchers Release FlashMap, a Flow Map Drafter for 6.14x Faster LLM Decoding
FlashMap is a speculative decoding drafter from Polymathic AI, Princeton and USC ISI researchers. It drafts 16 tokens jointly in one pass and reaches up to 6.14x lossless speedup on Qwen3-8B. The post Polymathic AI and Princeton Researchers Release FlashMap, a Flow Map Drafter for 6.14x Faster LLM Decoding appeared first on MarkTechPost .
