ConSA: Controllable Sparsity in Hybrid Attention via Learnable Allocation 文章

ArXiv CS.CL2026-06-17NEWSen作者: Yao Chen, Yinqi Yang, Junyuan Shang, Xiangzhao Hao, Simeng Zhang, Yilong Chen, Tingwen Liu, Shuohuan Wang, Dianhai Yu

ConSA: Controllable Sparsity in Hybrid Attention via Learnable Allocation · 相关技术