Component Ablation for Efficient Hybrid Language Model Architectures: Performance, Resilience, and Compression Implications 文章

ArXiv CS.AI2026-06-09NEWSen作者: Hector Borobia, Elies Segu\'i-Mas, Guillermina Tormo-Carb\'o
查看原文 →
关系图谱

Component Ablation for Efficient Hybrid Language Model Architectures: Performance, Resilience, and Compression Implications · 相关技术

相关技术

linear-attentionstate-spacelinear-time sequence mechanismssoftmax attention