TY  - RPRT
TI  - MoSKA: Mixture of Shared KV Attention for Efficient Long-Sequence LLM Inference
AU  - Myunghyun Rhee
AU  - Sookyung Choi
AU  - Euiseok Kim
AU  - Joonseop Sim
AU  - Youngpyo Joo
AU  - Hoshik Kim
PY  - 2025
DO  - 10.1109/lca.2025.3627539
UR  - https://arxiv.org/abs/2511.06010
ID  - 2511.06010
ER  -