When Correct Solutions Repeat: Rarity-Aware Credit Redistribution for GRPO 文章

ArXiv CS.AI2026-08-05PAPERen作者: Zhe Cao, Miaowen Wen, Fangjiong Chen
查看原文 →
关系图谱

When Correct Solutions Repeat: Rarity-Aware Credit Redistribution for GRPO · 相关技术

相关技术

Strategy CuesCue-GRPOFlow-GRPOReinforcement learning with verifiable rewards