Software Engineering Issue Resolution Training Data

SWE-Factory is an automated system for generating issue resolution training data and evaluation benchmarks. The dataset, created by Deep-Software-Analytics, was last updated on June 21, 2025. It is hosted on Hugging Face under the 'Swesetupbench Lite' repository.

Use Cases

• Training language models to generate code fixes based on issue descriptions and context.
• Benchmarking the performance of automated code repair systems on a standardized set of software issues.
• Analyzing patterns in software bugs and their resolutions using the provided issue and solution pairs.

Strengths

• Dataset is associated with a research paper (arXiv:2506.10954), indicating a documented methodology.
• Data is available in Parquet format, suitable for efficient processing with libraries like Polars and Pandas.

Limitations

• The exact number of rows, columns, and total size of the dataset is unknown.
• Sample data and specific column names are unavailable, making initial assessment difficult.

Provenance

Source

Deep-Software-Analytics via Hugging Face.

Collection Method

Generated by the SWE-Factory automated system for creating training data and benchmarks.

Freshness

Last updated on 2025-06-21.

Geography

Region tag suggests a US focus, but specific coverage is unknown.