PulseAugur
中
实时 19:11:06
English(EN) Channel Selection for Mixture-of-Experts Decoding Needs No Calibration Data

MoE解码优化无需校准数据,加速推理

研究人员开发了一种优化解码过程中Mixture-of-Experts (MoE)模型的新方法,无需校准数据。提出的“Prox”系统用top-k方法和权重派生读取器取代了传统的基于阈值的通道选择。该方法显著加速了MoE推理,速度提升高达2.69倍,同时准确性损失极小,并缓解了Gemma 4等模型中观察到的重复循环崩溃等问题。 AI

影响 这项研究可以显著加速大型MoE模型的推理速度,使其在实时应用中更加实用。

排序理由 学术论文,详细介绍了一种优化MoE模型的新方法。

在 Hugging Face Daily Papers 阅读 →

AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →

MoE解码优化无需校准数据,加速推理

本文如何被排名

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Research
学术论文,详细介绍了一种优化MoE模型的新方法。
Source corroboration
3 independent sources
Multiple independent publishers reporting the same story raises confidence that it's real and newsworthy.
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
26 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.
Coverage growth since scoring
+1 source(s) since last score
New sources have picked up this story since our last re-score. Score will update on the next scoring pass.

完整方法见我们的编辑标准。

报道来源 [3]

  1. arXiv cs.AI TIER_1 English(EN) · Geunhyoung Kim, Sung Ju Hwang ·

    Mixture-of-Experts解码的通道选择无需校准数据

    arXiv:2610.37386v1 Announce Type: cross Abstract: Mixture-of-Experts (MoE) models are bottlenecked by loading routed-expert weights from memory during interactive decoding at small batch sizes. Because most intermediate channels in a routed expert contribute little to any single …

  2. Hugging Face Daily Papers TIER_1 English(EN) ·

    Mixture-of-Experts解码的通道选择无需校准数据

    Mixture-of-Experts (MoE) models are bottlenecked by loading routed-expert weights from memory during interactive decoding at small batch sizes. Because most intermediate channels in a routed expert contribute little to any single token, channel selectors predict which channels ma…

  3. Hugging Face Daily Papers TIER_1 English(EN) ·

    Mixture-of-Experts解码的通道选择无需校准数据

    Prox replaces calibrated thresholds with top-k channel selection and a weight-derived reader to accelerate Mixture-of-Experts inference while preserving accuracy.