PGPO uses KL divergence to quantify token visual dependency and reshapes advantages in RLVR to amplify signals for visually grounded tokens, yielding 18.7% average gains on seven benchmarks.
Because the func- tion is globally strictly monotonically increasing across its domain[0,1], we obtain: IA > I B =⇒ω A > ω B (52)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Not All Tokens See Equally: Perception-Grounded Policy Optimization for Large Vision-Language Models
PGPO uses KL divergence to quantify token visual dependency and reshapes advantages in RLVR to amplify signals for visually grounded tokens, yielding 18.7% average gains on seven benchmarks.