GRPO's rank bias reinforces likely answers and neglects rare correct proofs; an unlikeliness reward that down-weights likely correct samples improves pass@N in formal theorem proving.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Rewarding the Unlikely: Lifting GRPO Beyond Distribution Sharpening
GRPO's rank bias reinforces likely answers and neglects rare correct proofs; an unlikeliness reward that down-weights likely correct samples improves pass@N in formal theorem proving.