arXiv cs.CL· Weihao Li, Zihang Xu, Zhongqiu Wang, Songwei Yu, Wenqiang Zhu, Yuechen Luo, Jingxuan Xu, Yanan Wu, Haoyang Huang, Ken Deng·· 2 天前AI 评分30
WEAVE：将真实世界代码演进织入依赖结构化任务，用于代码智能体的基准测试与训练
WEAVE: Weaving Real-World Code Evolution into Dependency-Structured Tasks for Benchmarking and Training Code Agents
AI 导读
WEAVE 框架将已验证的 PR 历史转化为可执行、依赖结构化的任务，通过仓库校准的类型化变更位点把相关 PR 组织成有界 DAG 并实例化为统一可执行任务。该框架还提出 Graph Credit，一种确定性的、基于测试的指标，用于衡量依赖一致的部分进展，并为强化学习提供稠密可验证奖励。实验表明 WEAVE 能生成连贯的可执行任务，比扁平评分更忠实地衡量进展，并为代码智能体训练提供有效监督。
正文 · 原文
Abstract:Real-world software engineering rarely consists of isolated changes. Features, migrations, and systematic fixes often unfold across multiple pull requests (PRs), requiring coordinated modifications across a repository. Yet most executable repository-level benchmarks retain a single issue or PR as the task unit, while existing multi-PR settings rely on coarse aggregation and structure-agnostic evaluation. This mismatch obscures dependencies among changes and limits progress measurement and learning signals for long-horizon code agents. We introduce WEAVE, a framework that transforms verified PR histories into executable, dependency-structured tasks. WEAVE leverages repository-calibrated typed change loci to organize related PRs into bounded DAGs and instantiate each as a unified executable task. We further introduce Graph Credit, a deterministic, test-grounded metric that measures dependency-consistent partial progress and supports dense, verifiable rewards for reinforcement learning. Experiments on task construction, agent evaluation, supervised fine-tuning, and reinforcement learning show that WEAVE yields coherent executable tasks, enables more faithful progress measurement than flat scoring, and provides effective supervision and verifiable rewards for training code agents.
| Comments: | 22 pages, including appendix |
| Subjects: | Software Engineering (cs.SE); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.22039 [cs.SE] |
| (or arXiv:2610.22039v1 [cs.SE] for this version) | |
| https://doi.org/10.48550/arXiv.2610.22039 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yuechen Luo [view email]
[v1]
Thu, 15 Oct 2026 15:12:56 UTC (8,746 KB)
来源：arXiv cs.CL · arxiv.org