# Data Contamination

> Data contamination is the presence of benchmark or test examples in a model's training data, which inflates its measured performance because it may have memorized the answers.

- Identifier: PTL-0178
- Category: Failure Modes & Evaluation
- Canonical URL: https://protologue.com/t/data-contamination/
- Also known as: benchmark contamination, test set contamination, training data leakage
- Introduced: 2023

## Description

Since models are trained on large web crawls, public benchmarks often leak into training sets. Contamination can be detected, for example, by prompting a model to complete the beginning of test instances and checking for verbatim matches. For application evaluations, the safeguard is a private, regularly refreshed test set.

## Related terms

- [Evals](https://protologue.com/t/evals/)
- [Knowledge Cutoff](https://protologue.com/t/knowledge-cutoff/)

## Sources

- Golchin and Surdeanu (2023). Time Travel in LLMs: Tracing Data Contamination in Large Language Models. https://arxiv.org/abs/2308.08493

## Cite this entry

Protologue. (2026). Data Contamination. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0178). https://protologue.com/t/data-contamination/

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)