CoT-BERT proposes a two-stage approach for sentence representation, leveraging Chain-of-Thought and contrastive learning to enhance unsupervised models like BERT.
