Dataset for Log Parsing: How Far Can Llama-2 and ChatGpt Go?

We present the results of our method on a new version of this benchmark, which has 2,000 log messages with manual annotations. we choose Loghub as it is a public resource of system logs for AI-based log analysis research. It contains logs from real and lab systems, without any changes or modifications. We use datasets from LogPAI, a research platform for log anomaly detection and analysis. LogPAI provides the most comprehensive collection of log datasets, covering different systems such as distributed, supercomputer, operating system, mobile, server and software systems.

本研究针对该基准测试集的新版本展示了我们所提方法的实验结果，该版本包含2000条带人工标注的日志消息。我们选取Loghub作为研究数据源，因其是面向基于AI的日志分析研究的公开系统日志资源。该资源包含来自真实系统与实验室系统的日志，未经过任何修改或调整。我们使用了来自LogPAI的数据集，LogPAI是面向日志异常检测与分析的研究平台。该平台提供了最为全面的日志数据集集合，涵盖分布式系统、超级计算机、操作系统、移动系统、服务器以及软件系统等多种类型的系统。