TY  - RPRT
TI  - Gated Sparse Attention: Combining Computational Efficiency with Training Stability for Long-Context Language Models
AU  - Alfred Shen
AU  - Aaron Shen
PY  - 2026
UR  - https://arxiv.org/abs/2601.15305
ID  - 2601.15305
ER  -