DeepSeek[1]
置顶
Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention
https://arxiv.org/abs/2502.11089 DeepSeek在AI领域的发展一直备受关注，其最新论文《Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention》更是引发了行
