The DOI matches its Crossref record; 23 of the 23 references it cites were found in a registry or an index.
Data Contamination: From Memorization to Exploitation
Open accessAbstract
Pretrained language models are typically trained on massive web-based datasets, which are often "contaminated" with downstream test sets. It is not clear to what extent models exploit the contaminated data for downstream tasks. We present a principled method to study this question. We pretrain BERT models on joint corpora of Wikipedia and labeled downstream datasets, and fine-tune them on the relevant task. Comparing performance between samples seen and unseen during pretraining enables us to define and quantify levels of memorization and exploitation. Experiments with two models and three downstream tasks show that exploitation exists in some cases, but in others the models memorize the contaminated data, but do not exploit it. We show that these two measures are affected by different factors such as the number of duplications of the contaminated data and the model size. Our results highlight the importance of analyzing massive web-scale datasets to verify that progress in NLP is obtained by better language understanding and not better data exploitation.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“Because data contamination is a well known issue in the field (Balloccu et al. [2024], Magar and Schwartz [2022], Sainz et al. [2023], Jacovi et al. [2023], Xu et al. [2024])”
From A Careful Examination of Large Language Model Performance on Grade School Arithmetic (2024) · cites this paper as Magar and Schwartz [2022]
In the citing paper
“Our framework intentionally combines 23 public/gated benchmarks with 14 private benchmarks to address training data contamination, where models may be optimized for public evaluation datasets27.”
From Holistic evaluation of large language models for medical tasks with MedHELM (2026) · cites this paper as [27]
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
The publisher hasn’t deposited this paper’s reference list.
Research works printed but not linked: 3
- Nicholas Carlini, Chang Liu, Úlfar Erlingsson, Jernej Kos, and Dawn Song. 2019. The secret sharer: Evaluating and testing unintended memorization in neural networks. In Proceedings of the 28th USENIX Conference on Security Symposium , SEC'19, page 267-284, USA. USENIX Association.
- Ilya Loshchilov and Frank Hutter. 2019. Decoupled weight decay regularization. In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019 . OpenReview.net.
- Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. Exploring the limits of transfer learning with a unified text-totext transformer. Journal of Machine Learning Research , 21(140):1-67.
Cited by
Sorted by citations
Towards a general-purpose foundation model for computational pathology
Chen, R. J., Ding, T., Lu, M. Y., Williamson, D. F. K., Jaume, G., Song, A. H., Chen, B., Zhang, A., Shao, D., Shaban, M., et al. (2024) · Nature Medicine · doi:10.1038/s41591-024-02857-3 · Cited by 1,560
Eight Things to Know about Large Language Models
Bowman, S. R. (2024) · Critical AI · doi:10.1215/2834703x-11556011 · Cited by 305
Fields, J., Chovanec, K., & Madiraju, P. (2024) · IEEE Access · doi:10.1109/access.2024.3349952 · Cited by 188
PromptRobust: Towards Evaluating the Robustness of Large Language Models on Adversarial Prompts
Zhu, K., Wang, J., Zhou, J., Wang, Z., Chen, H., Wang, Y., Yang, L., Ye, W., Zhang, Y., Gong, N., et al. (2023) · Proceedings of the 1st ACM Workshop on Large AI Systems and Models with Privacy and Safety Analysis · doi:10.1145/3689217.3690621 · Cited by 151
Speak, Memory: An Archaeology of Books Known to ChatGPT/GPT-4
Chang, K. K., Cramer, M. H., Soni, S., & Bamman, D. (2023) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2023.emnlp-main.453 · Cited by 176
Metadata
- Authors
- Magar, I., & Schwartz, R.
- ORCID
- Schwartz, Roy
- Venue
- Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)
- Published
- 2022
- Pages
- 157-165
- ISSN
- 0736-587X (ISSN-L)
- OpenAlex
- W4221145126
- Cited by
- 88
- References
- 0
- Citations by year
- 2026: 21 · 2025: 22 · 2024: 21 · 2023: 19 · 2022: 5
- FWCI
- 9.26
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- Proceedings of the conference - Association for Computational Linguistics. Meeting (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- Allen Institute for Artificial Intelligence
- Israel Science Foundation
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Keywords
- contaminated data, memorization, exploitation, pre-trained language models, BERT, Wikipedia, data duplication, model size, language comprehension
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
