we apply a staleness weight $w \left(\right. d_{t} \left.\right)$ that downweights tokens according to their age $d_{t}$ and drops them entirely once a threshold is exceeded
①数字：引入了指数衰减的陈旧度权重w(dt)。②金句：根据Token年龄降权并按阈值彻底丢弃。③非共识：传统RL常对整条轨迹统一计算优势，而此处针对长序列中早期生成Token的“过时”特性进行细粒度降权，是解决异步长序列采样偏离策略分布的精妙设计。