A causal framework for explaining the predictions of black-box sequence-to-sequence models.

David Alvarez-Melis, Tommi S. Jaakkola

Browse the full EMNLP paper archive.