研究人员推出 SMoES，一种用于引导混合专家（MoE）视觉语言模型（VLM）中专家路由的新颖方法。该方法利用动态软模态分数来考虑层依赖的融合模式，从而提高这些模型的有效性和效率。实验表明，SMoES 可以在多模态和语言任务上带来显著的提升，同时还能减少通信开销并提高实际部署中的吞吐量。 AI
影响 增强 MoE-VLM 的效率和有效性，可能提高多模态任务的性能。
排序理由 该集群描述了一篇详细介绍 MoE-VLM 新方法的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2604.23996v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) has become a prevalent backbone for large vision-language models (VLMs), yet how modality-specific signals should guide expert routing remains under-explored. Existing routing strategies are either hand-craf…
Mixture-of-Experts (MoE) has become a prevalent backbone for large vision-language models (VLMs), yet how modality-specific signals should guide expert routing remains under-explored. Existing routing strategies are either hand-crafted or modality-agnostic, relying on idealized p…