---
title: "Towards Data Contamination Detection for Modern Large"
description: "To get ground truth, they build an oracle by fine-tuning LLaMA-2 (70B) on known portions of six… 25 references, cited by 14 papers here."
authors: ["Vinay Samuel", "Yue Zhou", "Henry Peng Zou"]
year: 2025
venue: "Proceedings of the 31st International Conference on Computational Linguistics"
source: "https://aclanthology.org/2025.coling-main.338/"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/samuel-2025-data-contamination-detection-modern-large/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/samuel-2025-data-contamination-detection-modern-large/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2025](https://md.research.elamur.ai/free/en/papers/2025/page/6/) / Towards Data Contamination Detection for Modern Large Language Models: Limitations, Inconsistencies, and Oracle Challenges

# Towards Data Contamination Detection for Modern Large Language Models: Limitations, Inconsistencies, and Oracle Challenges

**Authors:** [Vinay Samuel](https://md.research.elamur.ai/free/en/authors/vinay-samuel/), [Yue Zhou](https://md.research.elamur.ai/free/en/authors/yue-zhou/) and [Henry Peng Zou](https://md.research.elamur.ai/free/en/authors/henry-peng-zou/) · **Year:** 2025 · **Venue:** [Proceedings of the 31st International Conference on Computational Linguistics](https://md.research.elamur.ai/free/en/venues/coling/) ([COLING 2025](https://md.research.elamur.ai/free/en/venues/coling/2025/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2025.coling-main.338/) · [PDF](https://aclanthology.org/2025.coling-main.338.pdf) · [OpenAlex](https://openalex.org/W4403679014)

**Citations:** 2 citations (OpenAlex, 2026-10-14) · [14](https://md.research.elamur.ai/free/en/papers/samuel-2025-data-contamination-detection-modern-large/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary

Data contamination detectors for LLMs have mostly been validated on older, widely circulated benchmarks and early models such as GPT-J. It has been unclear whether they work for current frontier models and harder test sets. The authors test five detectors: three ICLR methods (token completion overlap, Min-K% probability, canonical order statistical testing), a word perturbation quiz, and their own instance local order quiz. They apply them to four LLMs and eight benchmarks. To get ground truth, they build an oracle by fine-tuning LLaMA-2 (70B) on known portions of six benchmarks, with answers rewritten as chain-of-thought solutions.

The detectors rarely agree with one another. None of them tracks the known contamination level in the oracle, and each relies on assumptions that limit its practical use. This matters because benchmark gains of modern LLMs cannot yet be reliably separated from memorization, so contamination claims, whether positive or negative, should be treated with caution.

## Key points

- **Benchmarks:** six challenging ones (GSM8K, MMLU, BIG-Bench-Hard, ARC-Challenge, DROP, HumanEval) and two traditional ones (AGNews, IMDB).
- **Models:** GPT-4 (gpt-4-0613), Claude 3, LLaMA-3-Chat (70B) and LLaMA-2-Chat (70B). The paper names the Claude variant as Sonnet in the setup and Haiku in the results. Min-K% and canonical order testing need access to model parameters, so they could not be run on GPT-4 or Claude.
- **Detection on hard benchmarks:** no method flagged contamination in DROP. Only the authors' local order quiz partially flagged HumanEval (accuracy 0.36).
- **Oracle results:** the fine-tuning portions ranged from 0% to 100%, but no metric followed this trend. Counterintuitively, every Min-K% value dropped after fine-tuning.
- **Agreement between methods:** Spearman rank correlations across all experiments are weak at best. Examples include Min-K% vs. perturbation accuracy, and canonical order vs. the local order quiz, which hints that the latter could serve as a proxy for closed models. Code is available at github.com/vsamuel2003/data-contamination.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Carnegie Mellon University](https://md.research.elamur.ai/free/en/universities/carnegie-mellon-university/) — University, Pittsburgh, United States · 539 works · 179,533 citations
- [University of Illinois Chicago](https://md.research.elamur.ai/free/en/universities/university-of-illinois-chicago/) — University, Chicago, United States · 61 works · 20,107 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

## Cite

Cite as: Vinay Samuel, Yue Zhou and Henry Peng Zou. 2025. Towards Data Contamination Detection for Modern Large Language Models: Limitations, Inconsistencies, and Oracle Challenges. Proceedings of the 31st International Conference on Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{samuel2025data,
  title = {{Towards Data Contamination Detection for Modern Large Language Models: Limitations, Inconsistencies, and Oracle Challenges}},
  author = {Vinay Samuel, Yue Zhou and Henry Peng Zou},
  booktitle = {Proceedings of the 31st International Conference on Computational Linguistics},
  year = {2025},
  url = {https://aclanthology.org/2025.coling-main.338/}
}
```

- [← previous Theory of Mind in Large Language Models: Assessment and Enhancement](https://md.research.elamur.ai/free/en/papers/chen-2025-theory-mind-large-language-models/)
- [next → Towards Reasoning Ability of Small Language Models](https://md.research.elamur.ai/free/en/papers/srivastava-2025-towards-reasoning-ability-small-language/)
- [↑ back to papers of 2025](https://md.research.elamur.ai/free/en/papers/2025/page/6/)

## References

25 references, 16 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. AI@Meta. 2024. Llama 3 model card. [find](https://scholar.google.com/scholar?q=AI%40Meta%2E+2024%2E+Llama+3+model+card%2E)
2. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, et al. 2021. [Training verifiers to solve math word problems](https://md.research.elamur.ai/free/en/papers/cobbe-2021-training-verifiers-solve-math-word/). arXiv preprint arXiv:2110.14168.
3. Dheeru Dua, Yizhong Wang, Pradeep Dasigi, Gabriel Stanovsky, Sameer Singh, and Matt Gardner. 2019. [Drop: A reading comprehension benchmark requiring discrete reasoning over paragraphs](https://md.research.elamur.ai/free/en/papers/dua-2019-drop-reading-comprehension-benchmark-requiring/). arXiv preprint arXiv:1903.00161.
4. Shahriar Golchin and Mihai Surdeanu. 2024a. Data contamination quiz: A tool to detect and estimate contamination in large language models. Preprint, arXiv:2311.06233. [find](https://scholar.google.com/scholar?q=Shahriar+Golchin+and+Mihai+Surdeanu%2E+2024a%2E+Data+contamination+quiz%3A+A+tool+to+detect+and+estimate+contamination+in+large+language+models%2E+Preprint%2C+arXiv%3A2311%2E06233%2E)
5. Shahriar Golchin and Mihai Surdeanu. 2024b. [Time travel in llms: Tracing data contamination in large language models](https://md.research.elamur.ai/free/en/papers/golchin-2023-time-travel-llms-tracing-data/). Preprint, arXiv:2308.08493.
6. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2020. [Measuring massive multitask language understanding](https://md.research.elamur.ai/free/en/papers/hendrycks-2020-measuring-massive-multitask-language-understanding/). In International Conference on Learning Representations.
7. Simone Balloccu, Patrícia Schmidtová, Mateusz Lango, and Ondřej Dušek. 2024. [Leak, cheat, repeat: Data contamination and evaluation malpractices in closed-source llms](https://md.research.elamur.ai/free/en/papers/balloccu-2024-leak-cheat-repeat-data-contamination/). arXiv preprint arXiv:2402.03927.
8. Alon Jacovi, Avi Caciularu, Omer Goldman, and Yoav Goldberg. 2023. Stop uploading test data in plain text: Practical strategies for mitigating data contamination by evaluation benchmarks. arXiv preprint arXiv:2305.10160. [find](https://scholar.google.com/scholar?q=Alon+Jacovi%2C+Avi+Caciularu%2C+Omer+Goldman%2C+and+Yoav+Goldberg%2E+2023%2E+Stop+uploading+test+data+in+plain+text%3A+Practical+strategies+for+mitigating+data+contamination+by+evaluation+benchmarks%2E+arXiv+prepri)
9. Nicholas Carlini, Florian Tramer, Eric Wallace, Matthew Jagielski, Ariel Herbert-Voss, Katherine Lee, Adam Roberts, Tom Brown, Dawn Song, Ulfar Erlingsson, Alina Oprea, and Colin Raffel. 2021. [Extracting training data from large language models](https://md.research.elamur.ai/free/en/papers/carlini-2020-extracting-training-data-large-language/). Preprint, arXiv:2012.07805.
10. Percy Liang, Rishi Bommasani, Tony Lee, Dimitris Tsipras, Dilara Soylu, Michihiro Yasunaga, et al. 2023. Holistic evaluation of language models. Preprint, arXiv:2211.09110. [find](https://scholar.google.com/scholar?q=Percy+Liang%2C+Rishi+Bommasani%2C+Tony+Lee%2C+Dimitris+Tsipras%2C+Dilara+Soylu%2C+Michihiro+Yasunaga%2C+et+al%2E+2023%2E+Holistic+evaluation+of+language+models%2E+Preprint%2C+arXiv%3A2211%2E09110%2E)
11. Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, Alex Ray, Raul Puri, Gretchen Krueger, Michael Petrov, Heidy Khlaaf, Girish Sastry, Pamela Mishkin, Brooke Chan, Scott Gray, Nick Ryder, Mikhai… — [Evaluating Large Language Models Trained on Code](https://md.research.elamur.ai/free/en/papers/chen-2021-evaluating-large-language-models-trained/)
12. Peter Clark, Isaac Cowhey, Oren Etzioni, Tushar Khot, Ashish Sabharwal, Carissa Schoenick, and Oyvind Tafjord. 2018. [Think you have solved question answering? try arc, the ai2 reasoning challenge](https://md.research.elamur.ai/free/en/papers/clark-2018-think-have-solved-question-answering/). arXiv preprint arXiv:1803.05457.
13. Andrew Maas, Raymond E Daly, Peter T Pham, Dan Huang, Andrew Y Ng, and Christopher Potts. 2011. [Learning word vectors for sentiment analysis](https://md.research.elamur.ai/free/en/papers/maas-2011-learning-word-vectors-sentiment-analysis/). In Proceedings of the 49th annual meeting of the association for computational linguistics: Human language technologies, pages 142–150.
14. OpenAI. 2023. [Gpt-4 technical report](https://md.research.elamur.ai/free/en/papers/openai-2023-gpt-4-technical-report/). arXiv:2303.08774.
15. Preprint, Yonatan Oren, Nicole Meister, Niladri S. Chatterji, Faisal Ladhak, and Tatsunori Hashimoto. 2024. Proving test set contamination in black-box language models. In The Twelfth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Preprint%2C+Yonatan+Oren%2C+Nicole+Meister%2C+Niladri+S%2E+Chatterji%2C+Faisal+Ladhak%2C+and+Tatsunori+Hashimoto%2E+2024%2E+Proving+test+set+contamination+in+black%2Dbox+language+models%2E+In+The+Twelfth+International+Co)
16. Jack W. Rae, Sebastian Borgeaud, Trevor Cai, Katie Millican, Jordan Hoffmann, Francis Song, et al. 2022. Scaling language models: Methods, analysis and insights from training gopher. Preprint, arXiv:2112.11446. [find](https://scholar.google.com/scholar?q=Jack+W%2E+Rae%2C+Sebastian+Borgeaud%2C+Trevor+Cai%2C+Katie+Millican%2C+Jordan+Hoffmann%2C+Francis+Song%2C+et+al%2E+2022%2E+Scaling+language+models%3A+Methods%2C+analysis+and+insights+from+training+gopher%2E+Preprint%2C+arXiv%3A2)
17. Oscar Sainz, Jon Ander Campos, Iker García-Ferrero, Julen Etxaniz, Oier Lopez de Lacalle, and Eneko Agirre. 2023. [Nlp evaluation in trouble: On the need to measure llm data contamination for each benchmark](https://md.research.elamur.ai/free/en/papers/sainz-2023-nlp-evaluation-trouble-need-measure/). arXiv preprint arXiv:2310.18018.
18. Weijia Shi, Anirudh Ajith, Mengzhou Xia, Yangsibo Huang, Daogao Liu, Terra Blevins, Danqi Chen, and Luke Zettlemoyer. 2023. Detecting pretraining data from large language models. Preprint, arXiv:2310.16789. [find](https://scholar.google.com/scholar?q=Weijia+Shi%2C+Anirudh+Ajith%2C+Mengzhou+Xia%2C+Yangsibo+Huang%2C+Daogao+Liu%2C+Terra+Blevins%2C+Danqi+Chen%2C+and+Luke+Zettlemoyer%2E+2023%2E+Detecting+pretraining+data+from+large+language+models%2E+Preprint%2C+arXiv%3A2310%2E)
19. Reza Shokri, Marco Stronati, Congzheng Song, and Vitaly Shmatikov. 2017. [Membership inference attacks against machine learning models](https://md.research.elamur.ai/free/en/papers/shokri-2016-membership-inference-attacks-against-machine/). Preprint, arXiv:1610.05820.
20. Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao, Abu Awal Md Shoeb, Abubakar Abid, Adam Fisch, Adam R. Brown, et al. 2023. [Beyond the imitation game: Quantifying and extrapolating the capabilities of language models](https://md.research.elamur.ai/free/en/papers/srivastava-2022-imitation-game-quantifying-extrapolating-capabilities/). Transactions on Machine Learning Research.
21. Mirac Suzgun, Nathan Scales, Nathanael Schärli, Sebastian Gehrmann, Yi Tay, Hyung Won Chung, Aakanksha Chowdhery, Quoc V Le, Ed H Chi, Denny Zhou, , and Jason Wei. 2022. [Challenging big-bench tasks and whether chain-of-thought can solve them](https://md.research.elamur.ai/free/en/papers/suzgun-2023-challenging-big-bench-tasks-whether/). arXiv preprint arXiv:2210.09261.
22. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. 2023. [Llama 2: Open foundation and fine-tuned chat models](https://md.research.elamur.ai/free/en/papers/touvron-2023-llama-2-open-foundation-fine/). arXiv preprint arXiv:2307.09288.
23. Feng Yao, Yufan Zhuang, Zihao Sun, Sunan Xu, Animesh Kumar, and Jingbo Shang. 2024. Data contamination can cross language barriers. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 17864–17875, Miami, Florida, USA. Association for Computational Linguis… [find](https://scholar.google.com/scholar?q=Feng+Yao%2C+Yufan+Zhuang%2C+Zihao+Sun%2C+Sunan+Xu%2C+Animesh+Kumar%2C+and+Jingbo+Shang%2E+2024%2E+Data+contamination+can+cross+language+barriers%2E+In+Proceedings+of+the+2024+Conference+on+Empirical+Methods+in+Natura)
24. Xiang Zhang, Junbo Zhao, and Yann LeCun. 2015. [Character-level convolutional networks for text classification](https://md.research.elamur.ai/free/en/papers/zhang-2015-character-level-convolutional-networks-text/). Advances in neural information processing systems, 28.
25. Henry Zou, Vinay Samuel, Yue Zhou, Weizhi Zhang, Liancheng Fang, Zihe Song, Philip Yu, and Cornelia Caragea. 2024. ImplicitAVE: An open-source dataset and multimodal LLMs benchmark for implicit attribute value extraction. In Findings of the Association for Computational Linguistics ACL 2024, pages 3… [find](https://scholar.google.com/scholar?q=Henry+Zou%2C+Vinay+Samuel%2C+Yue+Zhou%2C+Weizhi+Zhang%2C+Liancheng+Fang%2C+Zihe+Song%2C+Philip+Yu%2C+and+Cornelia+Caragea%2E+2024%2E+ImplicitAVE%3A+An+open%2Dsource+dataset+and+multimodal+LLMs+benchmark+for+implicit+attrib)

*Updated 30 October 2026*