手法を比較

	自己教師ありTransformer	RoBERTaベースの分類
分野	深層学習	深層学習
系統	Machine learning	Machine learning
提唱年	2017–2019	2019
提唱者	Vaswani et al. (architecture); Devlin et al. (BERT self-supervised paradigm)	Liu, Y. et al. (Facebook AI Research / University of Washington)
種類	Self-supervised deep learning model	Pre-trained transformer fine-tuned for sequence classification
原典	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.	Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., & Stoyanov, V. (2019). RoBERTa: A Robustly Optimized BERT Pretraining Approach. arXiv preprint arXiv:1907.11692.
別名	SSL Transformer, self-supervised pretraining, masked self-attention pretraining, contrastive transformer	RoBERTa classifier, RoBERTa text classification, Robustly Optimized BERT Classification, RoBERTa fine-tuning for classification
関連	5	5
概要	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.	RoBERTa-based Classification applies the RoBERTa pre-trained transformer — trained more robustly than BERT with dynamic masking and larger batches — to text categorisation tasks by adding a lightweight classification head on top of the [CLS] token representation and fine-tuning the entire model on labelled examples. It consistently matches or outperforms BERT on standard NLP benchmarks.