Blockwise Advantage Estimation

Improve credit assignment in multi-objective RL by decomposing advantages into segment-specific values. Use Outcome-Conditioned Baselines to reduce cross-objective interference without expensive rollouts, enabling better training signals for multi-step completions with different reward functions per segment.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit e311d65f31

Frequently asked questions

npx get-skill add adu2021/blockwise-advantage-estimation