Laser Self Reward

Compute reasoning rewards from the model's own next-token probability distribution at solution end. Integrates verification-based feedback into single model without separate evaluator, enabling efficient RL training with minimal overhead.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit 4588bdb62f

Frequently asked questions

npx get-skill add adu2021/laser-self-reward