A snapshot of data from the publication: Competition-Level Code Generation with AlphaCode. This snapshot contains: (HumanEval.zip) Human manual marking of the HumanEval dataset, showing true positives and false positives. (APPS.zip) Human manual marking of the APPS dataset, showing true positives, false positives, and semantically correct but algorithmically slow solutions. (alphacode_viz.zip) The code for the https://alphacode.deepmind.com/ website, showing visualizations of the AlphaCode model, selected samples, and explanations. Note that it is easier to view this through the https://alphacode.deepmind.com/ website. (dm-code_contests) The CodeContests dataset (which can be also downloaded following the instructions on https://githu...