QiMeng-PRepair: Precise Code Repair via Edit-Aware Reward Optimization
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918432262324224 |
|---|---|
| author | Ke, Changxin Zhang, Rui Guo, Jiaming Wen, Yuanbo Ding, Li Wang, Shuo Zhu, Xuyuan Peng, Xiong Huang, Di Du, Zidong Hu, Xing Guo, Qi Chen, Yunji |
| author_facet | Ke, Changxin Zhang, Rui Guo, Jiaming Wen, Yuanbo Ding, Li Wang, Shuo Zhu, Xuyuan Peng, Xiong Huang, Di Du, Zidong Hu, Xing Guo, Qi Chen, Yunji |
| contents | Large Language Models (LLMs) achieve strong program repair performance but often suffer from over-editing, where excessive modifications overwrite correct code and hinder bug localization. We systematically quantify its impact and introduce precise repair task, which maximizes reuse of correct code while fixing only buggy parts. Building on this insight, we propose PRepair, a framework that mitigates over-editing and improves repair accuracy. PRepair has two components: Self-Breaking, which generates diverse buggy programs via controlled bug injection and min-max sampling, and Self-Repairing, which trains models with Edit-Aware Group Relative Policy Optimization (EA-GRPO) using an edit-aware reward to encourage minimal yet correct edits. Experiments show that PRepair improves repair precision by up to 31.4% under $\mathrm{fix}_1@1$, a metric that jointly considers repair correctness and extent, and significantly increases decoding throughput when combined with speculative editing, demonstrating its potential for precise and practical code repair. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_05963 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | QiMeng-PRepair: Precise Code Repair via Edit-Aware Reward Optimization Ke, Changxin Zhang, Rui Guo, Jiaming Wen, Yuanbo Ding, Li Wang, Shuo Zhu, Xuyuan Peng, Xiong Huang, Di Du, Zidong Hu, Xing Guo, Qi Chen, Yunji Software Engineering Machine Learning Large Language Models (LLMs) achieve strong program repair performance but often suffer from over-editing, where excessive modifications overwrite correct code and hinder bug localization. We systematically quantify its impact and introduce precise repair task, which maximizes reuse of correct code while fixing only buggy parts. Building on this insight, we propose PRepair, a framework that mitigates over-editing and improves repair accuracy. PRepair has two components: Self-Breaking, which generates diverse buggy programs via controlled bug injection and min-max sampling, and Self-Repairing, which trains models with Edit-Aware Group Relative Policy Optimization (EA-GRPO) using an edit-aware reward to encourage minimal yet correct edits. Experiments show that PRepair improves repair precision by up to 31.4% under $\mathrm{fix}_1@1$, a metric that jointly considers repair correctness and extent, and significantly increases decoding throughput when combined with speculative editing, demonstrating its potential for precise and practical code repair. |
| title | QiMeng-PRepair: Precise Code Repair via Edit-Aware Reward Optimization |
| topic | Software Engineering Machine Learning |
| url | https://arxiv.org/abs/2604.05963 |