AgentExecutor: Partial Code Execution via Agentic Context Generation
Source: https://arxiv.org/abs/2608.05959
PDF: https://arxiv.org/pdf/2608.05959v1
Content fetched: 2026-08-09T15:34:20.317851+00:00
Grounding: OpenCLI arXiv metadata and abstract; Obsidian evidence: OpenClaw定时任务/论文流水线/2026-08-09-论文流水线.md
Metadata
- Author(s): Junkai Chen, Chengran Yang, Xing Hu, Zhenhao Li, Xin Xia, David Lo
- Original date: 2026-08-06
- Platform: arxiv
- AAIF quality score: 4
中文摘要
AgentExecutor 处理残缺代码片段执行问题:多 agent 流程先准备执行环境,再通过动态探索迭代补上下文,最后用程序合成演化 prefix。论文在 Stack Overflow 片段和开源项目代码两类数据上评估,报告最高 94% 与 90% coverage,相对 Treefix 分别提升 19.9% 和 13.8%,同时执行时间最多降低 80.3%、成本最多降低 56.6%。
English Summary
AgentExecutor is a multi-agent framework for partial code execution. It prepares execution environments, dynamically explores and refines missing context, and evolves prefixes through program synthesis. The paper reports up to 94% and 90% code coverage on Stack Overflow snippets and open-source project code, outperforming Treefix while reducing time and cost.
Abstract excerpt
Executing code snippets is essential for dynamic program analysis, but it remains challenging to execute an arbitrary code snippet due to issues like missing context and incomplete dependencies. In this paper, we propose AgentExecutor, a novel multi-agent framework for partial code execution. Our approach introduces a three-phase design: execution environment preparation, dynamic exploration with iterative refinement, and prefix evolution via program synthesis. We evaluate AgentExecutor on two widely used datasets comprising Stack Overflow snippets and open-source project code. The results show that AgentExecutor achieves up to 94% and 90% code coverage, outperforming Treefix by 19.9% and 13.8%, respectively.
Why it matters for AAIF
残缺代码执行开始从静态补洞转向 agentic 环境准备、探索和 coverage-guided 优化。