On Leakage of Code Generation Evaluation Datasets

Alexandre Matton, Tom Sherborne, Dennis Aumiller +7

In this paper, we consider contamination by code generation test sets, in particular in their use in modern large language models. We discuss three possible sources of such contami…