Reconsidering Token Embeddings with the Definitions for Pre-trained Language Models

Open in new window