Gain Rl Angle Concentration

Improve RL training efficiency by 2.5× using angle concentration between token hidden states as a cost-effective data scheduling signal, selecting high-gradient samples dynamically.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit e477906da4

Frequently asked questions

npx get-skill add adu2021/gain-rl-angle-concentration