mbpp / README.md
albertvillanova's picture
Reorder split names (#2)
  - crowdsourced
  - expert-generated
  - crowdsourced
  - expert-generated
  - en
  - cc-by-4.0
  - monolingual
pretty_name: Mostly Basic Python Problems
  - n<1K
  - original
  - text2text-generation
task_ids: []
  - code-generation
  - config_name: full
      - name: task_id
        dtype: int32
      - name: text
        dtype: string
      - name: code
        dtype: string
      - name: test_list
        sequence: string
      - name: test_setup_code
        dtype: string
      - name: challenge_test_list
        sequence: string
      - name: train
        num_bytes: 176879
        num_examples: 374
      - name: test
        num_bytes: 244104
        num_examples: 500
      - name: validation
        num_bytes: 42405
        num_examples: 90
      - name: prompt
        num_bytes: 4550
        num_examples: 10
    download_size: 563743
    dataset_size: 467938
  - config_name: sanitized
      - name: source_file
        dtype: string
      - name: task_id
        dtype: int32
      - name: prompt
        dtype: string
      - name: code
        dtype: string
      - name: test_imports
        sequence: string
      - name: test_list
        sequence: string
      - name: train
        num_bytes: 63453
        num_examples: 120
      - name: test
        num_bytes: 132720
        num_examples: 257
      - name: validation
        num_bytes: 20050
        num_examples: 43
      - name: prompt
        num_bytes: 3407
        num_examples: 7
    download_size: 255053
    dataset_size: 219630

Dataset Card for Mostly Basic Python Problems (mbpp)

Table of Contents

Dataset Description

Dataset Summary

The benchmark consists of around 1,000 crowd-sourced Python programming problems, designed to be solvable by entry level programmers, covering programming fundamentals, standard library functionality, and so on. Each problem consists of a task description, code solution and 3 automated test cases. As described in the paper, a subset of the data has been hand-verified by us.

Released here as part of Program Synthesis with Large Language Models, Austin et. al., 2021.

Supported Tasks and Leaderboards

This dataset is used to evaluate code generations.


English - Python code

Dataset Structure

dataset_full = load_dataset("mbpp")
    test: Dataset({
        features: ['task_id', 'text', 'code', 'test_list', 'test_setup_code', 'challenge_test_list'],
        num_rows: 974

dataset_sanitized = load_dataset("mbpp", "sanitized")
    test: Dataset({
        features: ['source_file', 'task_id', 'prompt', 'code', 'test_imports', 'test_list'],
        num_rows: 427

Data Instances

mbpp - full

    'task_id': 1,
    'text': 'Write a function to find the minimum cost path to reach (m, n) from (0, 0) for the given cost matrix cost[][] and a position (m, n) in cost[][].',
    'code': 'R = 3\r\nC = 3\r\ndef min_cost(cost, m, n): \r\n\ttc = [[0 for x in range(C)] for x in range(R)] \r\n\ttc[0][0] = cost[0][0] \r\n\tfor i in range(1, m+1): \r\n\t\ttc[i][0] = tc[i-1][0] + cost[i][0] \r\n\tfor j in range(1, n+1): \r\n\t\ttc[0][j] = tc[0][j-1] + cost[0][j] \r\n\tfor i in range(1, m+1): \r\n\t\tfor j in range(1, n+1): \r\n\t\t\ttc[i][j] = min(tc[i-1][j-1], tc[i-1][j], tc[i][j-1]) + cost[i][j] \r\n\treturn tc[m][n]',
    'test_list': [
        'assert min_cost([[1, 2, 3], [4, 8, 2], [1, 5, 3]], 2, 2) == 8',
        'assert min_cost([[2, 3, 4], [5, 9, 3], [2, 6, 4]], 2, 2) == 12',
        'assert min_cost([[3, 4, 5], [6, 10, 4], [3, 7, 5]], 2, 2) == 16'],
    'test_setup_code': '',
    'challenge_test_list': []

mbpp - sanitized

    'source_file': 'Benchmark Questions Verification V2.ipynb',
    'task_id': 2,
    'prompt': 'Write a function to find the shared elements from the given two lists.',
    'code': 'def similar_elements(test_tup1, test_tup2):\n  res = tuple(set(test_tup1) & set(test_tup2))\n  return (res) ',
    'test_imports': [],
    'test_list': [
        'assert set(similar_elements((3, 4, 5, 6),(5, 7, 4, 10))) == set((4, 5))',
        'assert set(similar_elements((1, 2, 3, 4),(5, 4, 3, 7))) == set((3, 4))',
        'assert set(similar_elements((11, 12, 14, 13),(17, 15, 14, 13))) == set((13, 14))'

Data Fields

  • source_file: unknown
  • text/prompt: description of programming task
  • code: solution for programming task
  • test_setup_code/test_imports: necessary code imports to execute tests
  • test_list: list of tests to verify solution
  • challenge_test_list: list of more challenging test to further probe solution

Data Splits

There are two version of the dataset (full and sanitized), each with four splits:

  • train
  • evaluation
  • test
  • prompt

The prompt split corresponds to samples used for few-shot prompting and not for training.

Dataset Creation

See section 2.1 of original paper.

Curation Rationale

In order to evaluate code generation functions a set of simple programming tasks as well as solutions is necessary which this dataset provides.

Source Data

Initial Data Collection and Normalization

The dataset was manually created from scratch.

Who are the source language producers?

The dataset was created with an internal crowdsourcing effort at Google.


Annotation process

The full dataset was created first and a subset then underwent a second round to improve the task descriptions.

Who are the annotators?

The dataset was created with an internal crowdsourcing effort at Google.

Personal and Sensitive Information


Considerations for Using the Data

Make sure you execute generated Python code in a safe environment when evauating against this dataset as generated code could be harmful.

Social Impact of Dataset

With this dataset code generating models can be better evaluated which leads to fewer issues introduced when using such models.

Discussion of Biases

Other Known Limitations

Since the task descriptions might not be expressive enough to solve the task. The sanitized split aims at addressing this issue by having a second round of annotators improve the dataset.

Additional Information

Dataset Curators

Google Research

Licensing Information


Citation Information

  title={Program Synthesis with Large Language Models},
  author={Austin, Jacob and Odena, Augustus and Nye, Maxwell and Bosma, Maarten and Michalewski, Henryk and Dohan, David and Jiang, Ellen and Cai, Carrie and Terry, Michael and Le, Quoc and others},
  journal={arXiv preprint arXiv:2108.07732},


Thanks to @lvwerra for adding this dataset.