metadata

annotations_creators: []
language_creators:
  - crowdsourced
  - expert-generated
language:
  - code
license:
  - apache-2.0
multilinguality:
  - multilingual
pretty_name: HumanEval-X
size_categories:
  - unknown
source_datasets: []
task_categories:
  - sequence-modeling
task_ids:
  - language-modeling

HumanEval-X

Dataset Description

HumanEval-X is a benchmark for the evaluation of the multilingual ability of code generative models. It consists of 820 high-quality human-crafted data samples (each with test cases) in Python, C++, Java, JavaScript, and Go, and can be used for various tasks.

The dataset is currently used for two tasks: code generation and code translation. For code generation, the model uses declaration and docstring as input to generate the solution. For code translation, the model uses declarations in both languages and the solution in the source language as input, to generate solutions in the target language.

Languages

The dataset contains coding problems in 5 programming languages: Python, C++, Java, JavaScript, and Go.

Dataset Structure

To load the dataset you need to specify a subset among the 5 exiting languages [python, cpp, go, java, js]. By default python is loaded.

from datasets import load_dataset
load_dataset("loubnabnl/humaneval-x", "js")

DatasetDict({
    test: Dataset({
        features: ['task_id', 'prompt', 'declaration', 'canonical_solution', 'test', 'example_test'],
        num_rows: 164
    })
})

next(iter(data["train"]))
{'task_id': 'JavaScript/0',
 'prompt': '/* Check if in given list of numbers, are any two numbers closer to each other than\n  given threshold.\n  >>> hasCloseElements([1.0, 2.0, 3.0], 0.5)\n  false\n  >>> hasCloseElements([1.0, 2.8, 3.0, 4.0, 5.0, 2.0], 0.3)\n  true\n  */\nconst hasCloseElements = (numbers, threshold) => {\n',
 'declaration': '\nconst hasCloseElements = (numbers, threshold) => {\n',
 'canonical_solution': '  for (let i = 0; i < numbers.length; i++) {\n    for (let j = 0; j < numbers.length; j++) {\n      if (i != j) {\n        let distance = Math.abs(numbers[i] - numbers[j]);\n        if (distance < threshold) {\n          return true;\n        }\n      }\n    }\n  }\n  return false;\n}\n\n',
 'test': 'const testHasCloseElements = () => {\n  console.assert(hasCloseElements([1.0, 2.0, 3.9, 4.0, 5.0, 2.2], 0.3) === true)\n  console.assert(\n    hasCloseElements([1.0, 2.0, 3.9, 4.0, 5.0, 2.2], 0.05) === false\n  )\n  console.assert(hasCloseElements([1.0, 2.0, 5.9, 4.0, 5.0], 0.95) === true)\n  console.assert(hasCloseElements([1.0, 2.0, 5.9, 4.0, 5.0], 0.8) === false)\n  console.assert(hasCloseElements([1.0, 2.0, 3.0, 4.0, 5.0, 2.0], 0.1) === true)\n  console.assert(hasCloseElements([1.1, 2.2, 3.1, 4.1, 5.1], 1.0) === true)\n  console.assert(hasCloseElements([1.1, 2.2, 3.1, 4.1, 5.1], 0.5) === false)\n}\n\ntestHasCloseElements()\n',
 'example_test': 'const testHasCloseElements = () => {\n  console.assert(hasCloseElements([1.0, 2.0, 3.0], 0.5) === false)\n  console.assert(\n    hasCloseElements([1.0, 2.8, 3.0, 4.0, 5.0, 2.0], 0.3) === true\n  )\n}\ntestHasCloseElements()\n'}

Data Fields

task_id: indicates the target language and ID of the problem. Language is one of ["Python", "Java", "JavaScript", "CPP", "Go"].
prompt: the function declaration and docstring, used for code generation.
declaration: only the function declaration, used for code translation.
canonical_solution: human-crafted example solutions.
test: hidden test samples, used for evaluation.
example_test: public test samples (appeared in prompt), used for evaluation.

Data Splits

Each subset has one splits: test.

Citation Information

Refer to https://github.com/THUDM/CodeGeeX.