What AI says about Data Contamination Quiz (DCQ)

The DCQ project provides an open-source implementation of a data contamination detection tool for large language models, based on the paper "Data Contamination Quiz". It detects contamination by framing the task as a five-option multiple-choice quiz, where one option is the original dataset instance, three are word-level perturbations, and one is none of these; the model's ability to identify the original instance indicates prior exposure and enables estimation of contamination for a dataset partition. The repository hosts code, README, and accompanying resources (video, poster, media) under the Apache-2.0 license.