English(EN) Beyond the Previous Layer: Residual Predictive Structure in Sparse MoE Routing

新研究揭示稀疏 MoE 路由中的预测结构

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-17 04:00

研究人员在稀疏混合专家（MoE）模型的路由机制中发现了残差预测结构。通过分析冻结的 OLMoE 和 JetMoE 模型，他们发现，除了紧邻的前一层之外，还纳入早期层的专家选择，可以显著提高对下一路由器选择的预测能力。这种扩展的历史通过在线性和非线性解码实验中更高的 R^2 值证明，从而提高了预测准确性。 AI

影响识别出提高稀疏 MoE 模型路由效率的潜力。

排序理由学术论文，详细介绍了 AI 模型架构的新发现。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

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

本文如何被排名

Signal score: 11 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 学术论文，详细介绍了 AI 模型架构的新发现。[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, model release; 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: Same-day; Cluster formed today. Ranking reflects the current source set at time of score.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Hao Li, Yasuyuki Tahara, Yuichi Sei · 2026-09-17 04:00

超越上一层：稀疏MoE路由中的残差预测结构

arXiv:2609.17940v1 Announce Type: new Abstract: Sparse mixture-of-experts models route each token through a sequence of expert selections. We ask whether the immediately preceding selection adequately summarizes this trajectory for predicting the next router. Using frozen OLMoE a…

本文如何被排名

报道来源 [1]

超越上一层：稀疏MoE路由中的残差预测结构

相关实体

相关话题