无需微调的开源代码修复智能体，跨语言表现稳定且可重复。
Kozuchi Agent: A Language-Agnostic Open-Weight Agent for Software Repair
- 采用分阶段、状态持久化设计，支持跨语言自动修复。
- 在SWE-bench上解决374/500实例，Python任务排名第一。
- 无需闭源模型，适配多团队内部集群，操作简化至单点触发。
工业级软件工程团队亟需能将错误报告转化为正确补丁的LLM智能体，但大规模评估面临长时程、工具使用规范、上下文持久、异构集群及评估复用等挑战。本文提出Kozuchi Agent，一个语言无关的开源权重修复智能体与CI驱动的评估流水线。通过显式阶段划分、状态持久、确定性工具、模型无关动作接口及跨智能体测试时选择机制，实现运行可审计与可复现。基于本地部署的Qwen3.5-27B（无微调，TTS@8），在官方评测器上成功修复374/500 SWE-bench Verified实例。在Multi-SWE-bench Java任务中，相同270亿参数智能体修复41/128实例（32.03%），在严格开源提交中排名第一，总体排名第4；在Python任务中排名12/135，为开源系统第一。各阶段行为在不同语言间偏差不超过±5个百分点。剩余失败主要源于语义正确性不足、Java专用测试框架问题及选择错误。对比显示，其性能优于同参数量的开放与本地基线。候选多样性、选择遗憾与补丁可靠性分析表明，当前差距主因是语义正确性与选择策略，而非编辑格式或闭源模型访问。运维层面，可复用的CI阶段将操作触点从五次缩减至一次，兼容异构内部集群。
原文摘要 · Abstract (English)
Industrial software-engineering teams increasingly need LLM agents that turn bug reports into correct patches, yet benchmark-scale operation adds long horizons, tool-use discipline, context persistence, heterogeneous clusters, and evaluation reuse. We present Kozuchi Agent, a language-agnostic open-weight repair agent and CI-operated evaluation pipeline. Explicit phases, persistent state, deterministic tools, a model-independent action interface, and cross-agent test-time selection make runs auditable and repeatable. With locally hosted Qwen3.5-27B, no fine-tuning, and TTS@8, Kozuchi resolves 374/500 SWE-bench Verified instances on the official evaluator. Unchanged on Multi-SWE-bench Java, the same 27-billion-parameter agent resolves 41/128 instances (32.03%), ranking first among strict open-weight submissions and fourth of 42 overall; on Python it ranks 12th of 135 and first among open-weight systems. Per-phase behavior remains within +/-5 percentage points across languages. Remaining failures mainly reflect semantic correctness, Java-specific harness issues, and selection errors. Across both tracks, results compare favorably with open/local peers by parameter count. Analysis of candidate diversity, selector regret, and patch reliability shows that the remaining gap is primarily semantic correctness and selection rather than edit formatting or proprietary-model access. Operationally, reusable CI stages reduce operator touch-points from five to one across heterogeneous internal clusters.
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。