Hybrid Linear Attention Done Right: Efficient Distillation and Effective Architectures for Extremely Long Contexts Paper • 2601.22156 • Published Jan 29 • 14
