BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914563626106880 |
|---|---|
| author | Li, Yuanhao Wang, Hongbo Shang, Xiaotang Tang, Xunzhu Cao, Yiming Chen, Xuhong |
| author_facet | Li, Yuanhao Wang, Hongbo Shang, Xiaotang Tang, Xunzhu Cao, Yiming Chen, Xuhong |
| contents | Reinforcement learning for program repair is hindered by sparse execution feedback and coarse sequence-level rewards that obscure which edits actually fix bugs. We present BoostAPR, a three-stage framework addressing these challenges: (1) supervised fine-tuning on execution-verified demonstrations with reasoning traces, (2) training dual reward models--a sequence-level assessor and a line-level credit allocator--from execution outcomes, and (3) PPO optimization where the line-level model redistributes rewards to critical edit regions. This line-level credit assignment operates at an intermediate granularity naturally suited to code changes. Trained on SWE-Gym and evaluated on four benchmarks, BoostAPR achieves 40.7% on SWE-bench Verified (+22.9pp over base model), 24.8% on Defects4J (Python-to-Java transfer), 84.5% on HumanEval-Java, and 95.0% on QuixBugs, achieving competitive results among open-source models with strong cross-language generalization. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_09134 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models Li, Yuanhao Wang, Hongbo Shang, Xiaotang Tang, Xunzhu Cao, Yiming Chen, Xuhong Artificial Intelligence Software Engineering Reinforcement learning for program repair is hindered by sparse execution feedback and coarse sequence-level rewards that obscure which edits actually fix bugs. We present BoostAPR, a three-stage framework addressing these challenges: (1) supervised fine-tuning on execution-verified demonstrations with reasoning traces, (2) training dual reward models--a sequence-level assessor and a line-level credit allocator--from execution outcomes, and (3) PPO optimization where the line-level model redistributes rewards to critical edit regions. This line-level credit assignment operates at an intermediate granularity naturally suited to code changes. Trained on SWE-Gym and evaluated on four benchmarks, BoostAPR achieves 40.7% on SWE-bench Verified (+22.9pp over base model), 24.8% on Defects4J (Python-to-Java transfer), 84.5% on HumanEval-Java, and 95.0% on QuixBugs, achieving competitive results among open-source models with strong cross-language generalization. |
| title | BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models |
| topic | Artificial Intelligence Software Engineering |
| url | https://arxiv.org/abs/2605.09134 |