Are Structural Concepts Universal in Transformer Language Models? Towards Interpretable Cross-Lingual Generalization

Open in new window