1 paper
Yilong Wan, Yuqiang Li, Xianyi Wu
This paper investigates off-policy evaluation in contextual bandits, aiming to quantify the performance of a target policy using data collected under a different and potentially un…