Should You Mask 15% in Masked Language Modeling?
Open accessThe DOI matches its Crossref record; 22 of the 22 references it cites were found in a registry or an index.
Abstract
Masked language models (MLMs) conventionally mask 15% of tokens due to the belief that more masking would leave insufficient context to learn good representations; this masking rate has been widely used, regardless of model sizes or masking strategies.In this work, we revisit this important choice of MLM pre-training.We first establish that 15% is not universally optimal, and larger models should adopt a higher masking rate.Specifically, we find that masking 40% outperforms 15% for BERT-large size models on GLUE and SQuAD.Interestingly, an extremely high masking rate of 80% can still preserve 95% fine-tuning performance and most of the accuracy in linguistic probing, challenging the conventional wisdom about the role of the masking rate.We then examine the interplay between masking rates and masking strategies and find that uniform masking requires a higher masking rate compared to sophisticated masking strategies such as span or PMI masking.Finally, we argue that increasing the masking rate has two distinct effects: it leads to more corruption, which makes the prediction task harder; it also enables more predictions, which benefits optimization.Using this framework, we revisit BERT's 80-10-10 corruption strategy.Together, our results contribute to a better understanding of MLM pre-training. 1
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372
RoBERTa: A Robustly Optimized BERT Pretraining Approach
Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., & Stoyanov, V. (2019) · arXiv · doi:10.48550/arxiv.1907.11692 · Cited by 16,515
Self-citationExploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., Zhou, Y., Li, W., & Liu, P. J. (2019) · arXiv · doi:10.48550/arxiv.1910.10683 · Cited by 10,929
Lewis, M., Liu, Y., Goyal, N., Ghazvininejad, M., Mohamed, A., Levy, O., Stoyanov, V., & Zettlemoyer, L. (2020) · Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/2020.acl-main.703 · Cited by 8,821
Lewis, M., Liu, Y., Goyal, N., Ghazvininejad, M., Mohamed, A., Levy, O., Stoyanov, V., & Zettlemoyer, L. (2019) · arXiv · doi:10.48550/arxiv.1910.13461 · Cited by 8,816
Masked Autoencoders Are Scalable Vision Learners
He, K., Chen, X., Xie, S., Li, Y., Dollar, P., & Girshick, R. (2022) · 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52688.2022.01553 · Cited by 8,489
Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank
Socher, R., Perelygin, A., Wu, J., Chuang, J., Manning, C. D., Ng, A., & Potts, C. (2013) · Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d13-1170 · Cited by 6,878
SQuAD: 100,000+ Questions for Machine Comprehension of Text
Rajpurkar, P., Zhang, J., Lopyrev, K., & Liang, P. (2016) · Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d16-1264 · Cited by 6,208
ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
Lan, Z., Chen, M., Goodman, S., Gimpel, K., Sharma, P., & Soricut, R. (2019) · arXiv · doi:10.48550/arxiv.1909.11942 · Cited by 4,308
Cited by
Yang, L., Chen, H., Li, Z., Ding, X., & Wu, X. (2023) · arXiv · doi:10.48550/arxiv.2306.11489 · Cited by 239
Yang, L., Chen, H., Li, Z., Ding, X., & Wu, X. (2024) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2024.3360454 · Cited by 238
Language Model Behavior: A Comprehensive Survey
Chang, T. A., & Bergen, B. K. (2024) · Computational Linguistics · doi:10.1162/coli_a_00492 · Cited by 105
Findings of the BabyLM Challenge: Sample-Efficient Pretraining on Developmentally Plausible Corpora
Warstadt, A., Mueller, A., Choshen, L., Wilcox, E., Zhuang, C., Ciro, J., Mosquera, R., Paranjabe, B., Williams, A., Linzen, T., et al. (2023) · Proceedings of the BabyLM Challenge at the 27th Conference on Computational Natural Language Learning · doi:10.18653/v1/2023.conll-babylm.1 · Cited by 74
“Indeed, there are anecdotes of basic design choices for popular pipelines, such as the masking rate for BERT training (Wettig et al., 2023), being poorly tuned for years, despite hundreds or even thousands of papers using this training pipeline.”
The Plastic Surgery Hypothesis in the Era of Large Language Models
Xia, C. S., Ding, Y., & Zhang, L. (2023) · 2023 38th IEEE/ACM International Conference on Automated Software Engineering (ASE) · doi:10.1109/ase56229.2023.00047 · Cited by 67
Metadata
- Authors
- Wettig, A., Gao, T., Zhong, Z., & Chen, D.
- Venue
- Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics
- Published
- 2023
- Pages
- 2985-3000
- OpenAlex
- W4386566638
- Cited by
- 145
- References
- 22
- FWCI
- 25.14
- Type
- conference paper
- Language
- en
- Open access
- hybrid
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Should You Mask 15% in Masked Language Modeling? https://app.cochat.ai/papers/10.18653/v1/2023.eacl-main.217
- The paper
- Wettig, A., Gao, T., Zhong, Z., & Chen, D. (2023). Should You Mask 15% in Masked Language Modeling? Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics. https://doi.org/10.18653/v1/2023.eacl-main.217
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.