About

A decoupled context-compression co-processor utilizing gated cross-attention for linear-time, O(1)-memory sequence scaling in frozen large language models. (159 символов)

Stars

Watchers

Forks