F Grpo Divergence Alignment

Unify LLM alignment methods through f-divergence theory. f-GRPO extends GRPO to handle any divergence measure (KL, Jensen-Shannon, Hellinger), enabling tailored alignment objectives. f-HAL combines on-policy and off-policy preference learning to prevent reward hacking while maintaining safety alignment.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit 45f2b62b6e

Frequently asked questions

npx get-skill add adu2021/f-grpo-divergence-alignment