Analyzing Semantic Faithfulness of Language Models via Input Intervention on Question Answering

Open in new window