> ML_DATASET // MOSTLY-BASIC-PYTHON-PROBLEMS-MBPP_v1.0
MBPP: Mostly Basic Python Problems (Austin et al. 2021)
Google Research · Introductory Code Synthesis · 974 entry-level Python programming problems crowd-sourced with automated test cases
Introductory Code SynthesisCC-BY-4.0974 entry-level Python programming problems crowd-sourced with automated test casesopen
Dataset Profile & Characteristics
Label Type:English prompt description, Python code solution, and 3 test assertions
Languages:en, code
License Tier:permissive-open-source
Modalities:text, code
Intended Use
- Evaluating zero-shot and few-shot basic code synthesis in small to mid-scale LLMs
Prohibited / Discouraged Use
- Assessing enterprise system architecture skills
Bias, Leakage & Privacy Risk Analysis
Privacy / Sensitive Data Risks:
Crowdsourced elementary programming exercises; zero personal data.
Known Bias:
Elementary algorithms and data structure operations.
Known Benchmark Leakage:
Widespread memorization across open code models.