Learning in POMDPs is Sample-Efficient with Hindsight Observability.

Jonathan Lee, Alekh Agarwal, Christoph Dann, Tong Zhang

Browse the full ICML paper archive.