Skip to main content
Computer Science > Computation and Language
arXiv:1907.11692 (cs)
[Submitted on 26 Jul 2019]
RoBERTa: A Robustly Optimized BERT Pretraining Approach
Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, Veselin Stoyanov
View PDF
Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code.
Subjects:	Computation and Language (cs.CL)
Cite as:	arXiv:1907.11692 [cs.CL]
 	(or arXiv:1907.11692v1 [cs.CL] for this version)
 	
https://doi.org/10.48550/arXiv.1907.11692
Focus to learn more
Submission history
From: Myle Ott [view email]
[v1] Fri, 26 Jul 2019 17:48:29 UTC (45 KB)

Access Paper:
View PDFTeX Source
view license
Current browse context: cs.CL
< prev next >

newrecent2019-07
Change to browse by: cs
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
21 blog links (what is this?)
DBLP - CS Bibliography
listing | bibtex
Yinhan Liu
Myle Ott
Naman Goyal
Jingfei Du
Mandar Joshi
…
Export BibTeX Citation
Bookmark
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Demos
Related Papers
About arXivLabs
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)
About
Help
Contact
Subscribe
Copyright
Privacy Policy
Web Accessibility Assistance

arXiv Operational Status