Rlkv Cache Compression

Use reinforcement learning to identify reasoning-critical attention heads and apply selective KV cache compression, reducing cache by 20-50% while preserving reasoning quality. Maintains speedups up to 1.21x with minimal performance loss.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit d10654a7ee

Frequently asked questions

npx get-skill add adu2021/rlkv-cache-compression