Alvarez Melis, DavidJaakkola, Tommi S.2025-05-132017-11-14D. Alvarez-Melis and T. S. Jaakkola. "A causal framework for explaining the predictions of black-box sequence-to-sequence models". In: Proc. 2017 Conference on Empirical Methods in Natural Language Processing. EMNLP (Copenhagen, Denmark, Sept. 2017). Association for Computational Linguistics, 2017.https://dash.harvard.edu/handle/1/42717548We interpret the predictions of any black-box structured input-structured output model around a specific input-output pair. Our method returns an "explanation" consisting of groups of input-output tokens that are causally related. These dependencies are inferred by querying the black-box model with perturbed inputs, generating a graph over tokens from the responses, and solving a partitioning problem to select the most relevant components. We focus the general approach on sequence-to-sequence problems, adopting a variational autoencoder to yield meaningful input perturbations. We test our method across several NLP sequence generation tasks.EnglishConference Proceedings