标签
This paper introduces PAIR, a pairwise-aware inclusion reweighting method for adaptive rollout allocation in RLVR, improving sample efficiency and accuracy over pointwise allocators by correcting biases in pairwise gradient estimation.
TuneJury 是一个开源的成对奖励模型，用于文本到音乐生成，提供校准的偏好评分，并泛化到多个下游应用。