Counterfactuals As a Means for Evaluating Faithfulness of Attribution Methods in Autoregressive Language Models

Open in new window