Cornetto
收藏资源简介:
Cornetto是由苏黎世联邦理工学院开发的网络配置修复基准数据集,包含231个跨不同拓扑(20-754节点)和协议的复杂错误配置场景。该数据集通过语法生成和语义约束合成技术构建,确保配置的结构合理性和现实相关性,并采用差分数据平面分析验证功能正确性。其核心应用于评估大语言模型在网络自动化运维中的诊断与修复能力,为解决关键基础设施中配置错误的可靠修复提供标准化测试框架。
Cornetto is a benchmark dataset for network configuration repair developed by ETH Zurich. It contains 231 complex misconfiguration scenarios spanning diverse topologies (20–754 nodes) and network protocols. This dataset is constructed using syntax generation and semantic constraint synthesis technologies, ensuring the structural validity and real-world relevance of the configurations, and adopts differential data plane analysis to verify functional correctness. Its core application lies in evaluating the diagnostic and repair capabilities of large language models (LLMs) in network automated operation and maintenance, providing a standardized testing framework for reliable remediation of configuration errors in critical infrastructure.
- 1Benchmarking LLM-Driven Network Configuration Repair苏黎世联邦理工学院 · 2026年



