Dsdr Dual Scale Diversity Regularization

Improve LLM reasoning by promoting diversity at both trajectory and token levels simultaneously. Global (trajectory) scale rewards distinct correct solutions; local (token) scale applies entropy regularization per decision point. Dual-scale approach couples these via diversity-weighted allocation: solutions that are globally more distinctive receive stronger local regularization, focusing exploration where it matters most among underexplored correct modes.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit 5605b29d3f

Frequently asked questions

npx get-skill add adu2021/dsdr-dual-scale-diversity-regularization