MultiMend: Multilingual Program Repair with Context Augmentation and Multi-Hunk Patch Generation
Reza Gharibi, Mohammad Hadi Sadreddini, Seyed Mostafa Fakhrahmad · 2025
Read MultiMend: Multilingual Program Repair with Context Augmentation and Multi-Hunk Patch Generation by Reza Gharibi, Mohammad Hadi Sadreddini, Seyed Mostafa Fakhrahmad (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-16044.txt.
Description
Reza Gharibi and two co-authors present a 2025 arXiv system for automatically repairing buggy code across languages. Automated program repair has learned to patch single spots in single languages, yet real bugs sprawl across several locations and codebases differ. Their system, MultiMend, fine-tunes a code language model, then feeds it extra retrieved context lines beyond the faulty function.
A second mechanism assembles patches that span multiple repair sites, so one model can fix scattered bugs coherently. Across six benchmarks holding 5,501 bugs in four languages, MultiMend fixed 2,227 of them. Over two-thirds of those fixes matched the human developer's patch exactly, and 121 handled the hard multi-site cases. Ablations credited both the context retrieval and the multi-site assembly. The authors pitch the system as a practical step toward repair tools that work wherever code is written.
Catalog record
| CREATOR | Reza Gharibi, Mohammad Hadi Sadreddini, Seyed Mostafa Fakhrahmad |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 92 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-16044