Off-policy data in reinforcement learning with verifiable rewards (RLVR) can arise from asynchronous generation, rollout-training mismatch, and response reus…
机构:UIUC
来源:arXiv 2610.18738 | AI4Papers 论文推荐平台