方法对比

	自监督 GRU	自监督Transformer
领域	深度学习	深度学习
方法族	Machine learning	Machine learning
起源年份	2014–2019	2017–2019
提出者	Cho, K. et al. (GRU); self-supervised training paradigm from broader SSL literature	Vaswani et al. (architecture); Devlin et al. (BERT self-supervised paradigm)
类型	Self-supervised sequence model	Self-supervised deep learning model
开创性文献	Cho, K., van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation. In Proceedings of EMNLP 2014.	Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. Proceedings of NAACL-HLT 2019, 4171–4186.
别名	SS-GRU, Self-supervised Gated Recurrent Unit, GRU with self-supervised pretraining, Unsupervised GRU pretraining	SSL Transformer, self-supervised pretraining, masked self-attention pretraining, contrastive transformer
相关	4	5
摘要	Self-supervised GRU trains a Gated Recurrent Unit network using automatically constructed supervision signals — such as next-step prediction or masked token recovery — derived from the unlabeled data itself. The learned sequence representations are then fine-tuned on small labeled datasets, making high-quality sequential modeling feasible when annotations are scarce.	A self-supervised Transformer is a Transformer network pretrained using automatically constructed supervision signals — such as masked token prediction or next-sentence prediction — rather than human-annotated labels. The resulting representations are then fine-tuned or probed on downstream tasks. BERT, GPT, and ViT (Vision Transformer in masked-image modeling mode) are the most widely known instantiations of this paradigm.