A new evaluation mechanism and a learning strategy for de-biased and interpretable NLI models. Models co-learn sentence classification and evidence retrieval for the classification.
Primary LanguageJupyter NotebookMIT LicenseMIT
No issues in this repository yet.