MultiPL-E: A Scalable and Extensible Approach to Benchmarking Neural Code Generation
https://arxiv.org/abs/2208.08227