ResearchDatasetsCODEAGENTBENCHFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsRepo-level code generationCODEAGENTBENCH37.6Pass@138Code GenerationCODEAGENTBENCH NumpyML hard5Solved Problems5Code GenerationCODEAGENTBENCH NumpyML easy17Solved Problems5