F-GRPO: Don't Let Your Policy Learn the Obvious and Forget the Rare 事件

F-GRPO: Don't Let Your Policy Learn the Obvious and Forget the Rare arXiv:2602.06717v2 Announce Type: replace-cross Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) is commonly based on group sampling to estimate advantages and stabilize policy updates. In practice, computational limits often rule out very large groups, so training proceeds with finite rollout sets that can reinforce only the correct behavior they expose. At practical group sizes, updates can miss rare-correct tr