PulseAugur
中
实时 07:47:00
English(EN) SpecFit: Not All Draft Tokens Cost the Same in Offloaded MoE Inference

SpecFit 通过降低专家传输成本优化 MoE 推理速度

研究人员开发了 SpecFit,一种优化专家混合(MoE)模型推理速度的新颖方法,特别是在其专家权重被卸载到主机内存时。SpecFit 专注于通过选择最小化专家在主机和设备之间传输成本的草稿令牌来最大化预测吞吐量。该方法估计预期的已提交令牌,并根据激活的专家预测周期时间,确保每个令牌仅贡献新激活专家的成本。在 CNN/DailyMail 摘要和 GSM8K 上使用 Qwen3-30B-A3B 模型进行的实验表明,SpecFit 在解码速度上显著优于 EAGLE-3 等现有方法。 AI

影响 提高了专家混合模型的推理效率,可能带来更快、更具成本效益的 AI 部署。

排序理由 该集群包含一篇详细介绍优化 AI 模型推理新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

SpecFit 通过降低专家传输成本优化 MoE 推理速度

本文如何被排名

Signal score
20 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
该集群包含一篇详细介绍优化 AI 模型推理新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
paper, infra
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

  1. arXiv cs.AI TIER_1 English(EN) · Hongliang Cao, Xiuyu Li, Siqi Wang, Ziming Wang, Jingmin Zhu, Hailong Yang ·

    SpecFit:在卸载MoE推理中并非所有草稿令牌的成本都相同

    arXiv:2610.15417v1 Announce Type: new Abstract: Mixture-of-experts (MoE) models activate only a few experts per token, but when their expert weights are offloaded to host memory, host-to-device (H2D) expert transfers enter the critical path of decoding. Speculative decoding verif…