构建可组合且自适应的长周期编程代理框架,提升复杂任务执行效率。
openJiuwen: Beyond Static Harnesses for Long-Horizon Coding Agents

- 基于Rail机制实现单代理、子代理与协同流的能力组合
- 在SWE-bench Verified和Terminal-Bench 2.1上分别达82.6%和87.19%准确率
- 适合需要动态调整的复杂编程任务与多代理协作场景
长周期编程代理在不断演化的代码库状态中运行,日益依赖异构能力、委派代理与多代理协作。这带来两大互补挑战:一是开发者需在不重复构建调度逻辑的前提下,组合能力、重构执行流程并扩展复杂代理系统;二是复杂编程任务持续生成新证据(如语义诊断、执行结果、任务进度与上下文相关性变化),这些应动态影响后续运行决策。我们将其归纳为结构可组合性与运行时自适应性。本文提出openJiuwen,一个开源代理框架,兼顾开发者可组合性与任务执行自适应性。它提供共享执行底座与跨单代理、委派子代理及Swarm Flow的Rail-based能力组合,使开发者能在统一执行语义下构建复杂代理架构。同时,在固定模型策略下动态调整框架控制的运行时决策,让演化中的证据影响上下文、反馈与任务控制,以达成成功完成。我们在SWE-bench Verified和Terminal-Bench 2.1上系统评估,分别取得82.6%和87.19%的准确率,超越最强官方排行榜点估计3.4和3.39个百分点。结果表明,openJiuwen在复杂编程任务上兼具高性能与可组合、自适应的设计优势。
原文摘要 · Abstract (English)
Long-horizon coding agents operate over evolving repository states while increasingly relying on heterogeneous capabilities, delegated agents, and multi-agent coordination. These trends pose two complementary challenges for the agent harness. First, developers need to compose capabilities, reconfigure execution logic, and scale increasingly complex agent systems without repeatedly rebuilding orchestration. Second, complex coding tasks continuously produce new evidence---such as semantic diagnostics, execution outcomes, task progress, and changing context relevance---that should dynamically influence subsequent runtime decisions. We characterize these challenges as Structural Composability and Runtime Adaptivity. We present openJiuwen, an open-source harness designed for both developer composability and adaptive task execution. openJiuwen provides a shared execution substrate and Rail-based capability composition across single agents, delegated sub-agents, and Swarm Flow, enabling developers to construct sophisticated agent harnesses under common execution semantics. It further adapts framework-controlled runtime decisions around a fixed model policy, allowing evolving evidence to dynamically affect context, feedback, and task control toward successful completion. We systematically evaluate openJiuwen on SWE-bench Verified and Terminal-Bench 2.1, where it achieves 82.6% and 87.19%, respectively, exceeding the strongest selected official-leaderboard point estimates by 3.4 and 3.39 percentage points. These results show that openJiuwen achieves strong performance on complex coding tasks while providing a composable and adaptive harness design.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品;中文卡片由大模型生成,请以原文为准。