Detecting Benchmark Contamination Through Watermarking

Benchmark contamination poses a significant challenge to the reliability of Large Language Models (LLMs) evaluations, as it is difficult to assert whether a model has been trained on a test set. We introduce a solution to this problem by watermarking benchmarks before their release.The embedding involves reformulating the original questions with a watermarked LLM, in a way that does not alter the benchmark quality and utility. During evaluation, we can detect ``radioactivity'', i.e. traces that the text watermarks leave in the model during training, using a theoretically grounded statistical test. We test our method by pre-training 1B models from scratch on 10B tokens with controlled benchmark contamination, and validate its effectiveness in detecting contamination on ARC-Easy, ARC-Challenge, and MMLU. Results show similar benchmark utility post-rephrasing and successful contamination detection when models are contaminated enough to enhance performance, e.g. p-val =$10^{-3}$ for +5% on ARC-Easy.