Expert-based reward function training: the novel method to train sequence generators.

Joji Toyama, Yusuke Iwasawa, Kotaro Nakayama, Yutaka Matsuo

Browse the full ICLR paper archive.