Pepo Token Level Multimodal Policy

Replace uniform token-level advantages with perception-exploration gating that weights tokens by visual grounding strength. Adds 3.67 points to geometry reasoning and 5.32 to few-shot classification with <1% compute overhead. Works best for multimodal CoT where visual grounding anchors reasoning. Trigger: When using token-level RL on VLMs and want to emphasize visually-grounded reasoning steps.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit 190c79dc39

Frequently asked questions

npx get-skill add adu2021/pepo-token-level-multimodal-policy