Definition. Rollout group [shao2024deepseekmath, Section 4.1.2] [ftip-004B]
✍️source

    Fix a prompt \(x\), group size \(g\geq 2\), and behavior policy \(\pi _b\). A rollout group is the indexed family

    \[ \mathbf Y_x=(Y^{(1)},\ldots ,Y^{(g)}), \qquad Y^{(i)}\sim \pi _b(\mathord \cdot \mid x). \]

    DeepSeekMath samples several outputs for the same prompt so their rewards can define a group-relative baseline. The producing policy, decoder, group size, and conditional sampling dependence are part of the rollout law.