Extracting knowledge graphs from language models as a diagnostic benchmark of model performance (NeurIPS XAI 2021). - View it on GitHub