Optimal Sequential Annotations for Off-Policy Evaluation
It is shown how a limited budget for ground-truth data-annotation can be used via doubly-robust OPE with missing rewards, and the optimal annotation probabilities for sequential forward-monotone annotation protocols are characterized, and a feasible batch-adaptive implementation is provided.
Woojin Chae, Ezinne Nwankwo, Hai-Tong Qin et al.
· 0 citations