评估分布式系统中智能体代码修复能力的新基准 DDBench

AI 摘要

DDBench 是一个针对分布式系统代码修复的基准测试，包含从 13 个开源分布式系统中挖掘的 60 个历史 bug，分为三个难度等级。研究在仅症状和上下文增强两种条件下评估了 10 个 LLM，发现分布式调试能区分模型能力，且调试上下文能提升通过率，但对不同模型影响不对称。

正文节选

Evaluating Agentic Code Repair Capabilities in Distributed Systems Abstract LLM-based coding agents have advanced rapidly on single-process SWE tasks, with frontier models now clustering in the high-70s on SWE-bench Verified. Distributed-system debugging, however, remains an under-explored regime: bugs span processes, nodes, and protocol interactions, with root causes rarely recoverable from source alone and brute-force exploration intractable across non-deterministic interleavings. This leaves

发布时间：2026-08-18 12:00

抓取时间：2026-08-18 12:23

来源机构：arXiv