研究人员开发了QK-Normed MLA，一种无需完全缓存键即可稳定大型语言模型中注意力机制的方法。该技术通过分解RMSNorm并将静态权重吸收到现有投影中，将QK归一化集成到多头潜在注意力（MLA）中。与QK剪枝相比，该方法在保持MLA高效解码的同时，实现了更低的训练损失和更高的下游准确性，并且在Nvidia H800硬件上具有最小的延迟开销。 AI
影响 通过稳定注意力机制，能够更高效地训练和推理大型语言模型。
排序理由 该集群包含一篇详细介绍LLM新技术的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2606.16310v1 Announce Type: cross Abstract: Query-key (QK) normalization stabilizes attention by controlling the scale of queries and keys before the dot product, but is not immediately compatible with Multi-head Latent Attention (MLA). MLA achieves efficient decoding by ca…
Query-key (QK) normalization stabilizes attention by controlling the scale of queries and keys before the dot product, but is not immediately compatible with Multi-head Latent Attention (MLA). MLA achieves efficient decoding by caching low-dimensional latent states instead of ful…