1 paper · 1 filter
Zecheng Tang, Quantong Qiu, Yi Yang +6
The quadratic complexity of standard attention mechanisms poses a significant scalability bottleneck for large language models (LLMs) in long-context scenarios. While hybrid attent…