CodeXGLUE: A Machine Learning Benchmark Dataset for Code Understanding and Generation

Benchmark Model Rank Results
code-completion-on-codexglue-github-javaCodeGPT-adapted#1Accuracy (token-level): 77.13EM (line-level): 26.43…
code-completion-on-codexglue-py150CodeGPT-adapted#1Accuracy (token-level): 75.11EM (line-level): 39.65…
code-search-on-codexglue-advtestCodeBERT#3MRR: 27.19