“The issue of models being trained on benchmark data, leading to unreliable evaluation results.”

“A problem in AI training where test data or very similar problems are seen by the model.”

“An issue in AI benchmarks where test data might have been inadvertently included in training data, skewing results.”