---
source: arxiv
url: https://arxiv.org/abs/2607.09123v1
published_at: '2026-07-10T06:25:18'
authors:
- Quanjun Zhang
- Yi Zheng
- Ye Shang
- Weifeng Sun
- Haichuan Hu
- Chunrong Fang
- Zhenyu Chen
- Liang Xiao
topics:
- code-intelligence
- automated-testing
- multi-agent-software-engineering
- bug-reproduction
- repository-scale-reasoning
relevance_score: 0.96
run_id: materialize-outputs
language_code: zh-CN
---

# ReProAgent: Tool-Augmented Multi-Stage Agentic Generation of Bug Reproduction Tests from Issue Reports

## Summary
## 摘要
ReProAgent 将分阶段智能体、代码图检索、代码仓库探索和运行时验证结合起来，根据自然语言问题报告生成可执行的缺陷复现测试。它在 SWT-bench-lite 上复现了 58.43% 的问题，在 SWT-bench-verified 上复现了 70.30% 的问题，平均每个问题的成本为 $0.14。

## 问题

## 方法

## 结果

## Problem

## Approach

## Results

## Link
- [https://arxiv.org/abs/2607.09123v1](https://arxiv.org/abs/2607.09123v1)