barty/D4J-Repair

Dataset Description

• —Repository: MORepair
• —Paper: MORepair: Teaching LLMs to Repair Code via Multi-Objective Fine-tuning
• —Point of Contact: Boyang Yang

Dataset Summary

D4J-Repair is a curated subset of Defects4J, containing 371 single-function Java bugs from real-world projects. Each example includes a buggy implementation, its corresponding fixed version, and unit tests for verification.

Supported Tasks

• —Program Repair: Fixing bugs in Java functions
• —Code Generation: Generating correct implementations from buggy code

Dataset Structure

Each row contains:

• —task_id: Unique identifier for the task (in format: projectname-bugid)
• —buggy_code: The buggy implementation
• —fixed_code: The correct implementation
• —file_path: Original file path in the Defects4J repository
• —issue_title: Title of the bug
• —issue_description: Description of the bug
• —start_line: Start line of the buggy function
• —end_line: End line of the buggy function

Source Data

This dataset is derived from Defects4J, a collection of reproducible bugs from real-world Java projects. We carefully selected and processed single-function bugs to create this benchmark.

Citation

@article{morepair,
author = {Yang, Boyang and Tian, Haoye and Ren, Jiadong and Zhang, Hongyu and Klein, Jacques and Bissyande, Tegawende and Le Goues, Claire and Jin, Shunfu},
title = {MORepair: Teaching LLMs to Repair Code via Multi-Objective Fine-Tuning},
year = {2025},
publisher = {Association for Computing Machinery},
issn = {1049-331X},
url = {https://doi.org/10.1145/3735129},
doi = {10.1145/3735129},
journal = {ACM Trans. Softw. Eng. Methodol.},
}