English(EN) Ban&Pick: Enhancing Performance and Efficiency of MoE-LLMs via Smarter Routing

Ban&Pick 策略提升 MoE-LLM 性能和推理速度

作者 PulseAugur 编辑部 · [1 个来源] · 2026-08-27 04:00

研究人员开发了一种名为 Ban&Pick 的训练后策略，以提高专家混合（MoE）大语言模型的性能和效率。该方法解决了关键专家利用不足以及固定专家数量导致冗余的问题。通过识别和加强高影响力专家，Ban&Pick 提高了准确性，同时动态剪枝不太关键的专家则加快了推理速度。该策略在 DeepSeek 和 Qwen3 等模型上，在各种基准测试中都显示出显著的提升，且无需重新训练或架构更改。 AI

影响提高了 MoE-LLM 的效率和准确性，可能加速这些模型的采用。

排序理由该集群描述了一篇详细介绍改进 MoE-LLM 新方法的最新研究论文。[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: 25 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群描述了一篇详细介绍改进 MoE-LLM 新方法的最新研究论文。[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, 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: 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]

arXiv cs.LG TIER_1 English(EN) · Yuanteng Chen, Peisong Wang, Yuantian Shao, Nanxin Zeng, Chang Xu, Jian Cheng · 2026-08-27 04:00

Ban&Pick：通过更智能的路由增强 MoE-LLMs 的性能和效率

arXiv:2509.06346v3 Announce Type: replace Abstract: Sparse Mixture-of-Experts (MoE) has become a key architecture for scaling large language models (LLMs) efficiently. Recent fine-grained MoE designs introduce hundreds of experts per layer, with multiple experts activated per tok…

本文如何被排名

报道来源 [1]

Ban&Pick：通过更智能的路由增强 MoE-LLMs 的性能和效率

相关实体

相关话题